Multimedia Comm03

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/344445284
Multimedia Communications (Revised)
Book · October 2020

DOI: 10.13140/RG.2.2.14132.27529
CITATIONS READS
0 317
1 author:
Terrence Patrick McGarty

Massachusetts Institute of Technology
230 PUBLICATIONS 741 CITATIONS
SEE PROFILE
All content following this page was uploaded by Terrence Patrick McGarty on 01 October 2020.
The user has requested enhancement of the downloaded file.

TERRENCE MCGARTY
MULTIMEDIA
COMMUNICATIONS
Taught at MIT Fall 1989, Revised 2021
Multimedia Communications
Terrence P. McGarty Ph.D.

Massachusetts Institute of Technology
Fall 1989
© Copyright 1989, 2021 by Terrence P. McGarty, All Rights Reserved, updated 2021.
2 |Page
Table of Contents
1 Introduction ......................................................................................................................... 13
1.1 Definition of Multimedia ............................................................................................ 13
1.1.1 The Image ............................................................................................................... 15
1.1.2 The User .................................................................................................................. 15
1.1.3 Interaction ............................................................................................................... 15
1.2 Current Systems .......................................................................................................... 16
1.2.1 Computers ............................................................................................................... 18
1.2.2 Presentation Formats ............................................................................................... 19
1.2.3 Interaction Formats ................................................................................................. 20
1.3 The Paradigm .............................................................................................................. 21
1.3.1 Individual User........................................................................................................ 22
1.3.2 Physical User .......................................................................................................... 23
1.3.3 SubSession .............................................................................................................. 26
1.3.4 Session .................................................................................................................... 26
1.4 Elements of Multimedia Systems ............................................................................... 27
1.4.1 Display .................................................................................................................... 30
1.4.2 Applications Software ............................................................................................. 31
1.4.3 1.4.3 Services .......................................................................................................... 33
1.4.4 1.4.4 Network Software .......................................................................................... 34
1.4.5 1.4.5 Storage ........................................................................................................... 34
1.5 Key Problems ............................................................................................................. 34
1.5.1 Problem 1: Characterization ................................................................................... 35
1.5.2 Problem 2: Utilization ............................................................................................. 35
1.6 Processor/ Packetizer .................................................................................................. 36
1.7 Outline of Book .......................................................................................................... 38
2 Multimedia Communications ............................................................................................ 41
2.1 Current Definitions of Multimedia ............................................................................. 41
2.2 Philosophical Pillars ................................................................................................... 43
2.3 Multimedia; Elements And Structure ......................................................................... 50
3 |Page
2.3.1 Elements of Multimedia .......................................................................................... 50
2.3.2 Value Creation in Multimedia ................................................................................ 51
2.3.3 The Multimedia Food Chain and Its Missing Links ............................................... 52
2.3.4 Elements and Structures .......................................................................................... 52
2.4 The Messenger; Communications And Infrastructure................................................ 63
2.4.1 Communications Systems ....................................................................................... 64
2.4.2 Infrastructures ......................................................................................................... 66
2.4.3 Current Infrastructure Options ................................................................................ 69
2.5 Deconstruction ............................................................................................................ 73
2.6 Paradigms and World Views ...................................................................................... 73
2.7 Questions .................................................................................................................... 80
2.8 Hermeneutics .............................................................................................................. 87
2.8.1 Hermeneutic Principles ........................................................................................... 87
2.8.2 Hermeneutic Methodology Applied ....................................................................... 91
2.8.3 Communications at the Conversational Layer: Hermeneutics Applied .................. 92
2.9 Semiotics .................................................................................................................... 99
2.9.1 A Definition ............................................................................................................ 99
2.9.2 Semiotic Application ............................................................................................ 102
2.9.3 The Multimedia Database; A Semiotic Example.................................................. 103
2.10 Conclusions And Observations ................................................................................ 105
3 MultiMedia Elements ....................................................................................................... 107
3.1 The Media Construct ................................................................................................ 107
3.2 Images ....................................................................................................................... 110
3.3 Means of Characterization ........................................................................................ 111
3.3.1 Achromatic Images ............................................................................................... 114
3.4 Classic Color Theory ................................................................................................ 120
3.5 Spectra and Measurements ....................................................................................... 127
3.5.1 Classic Spectrometer ............................................................................................. 127
3.5.2 Fourier Transform Spectrometer........................................................................... 128
3.5.3 Video Means of Characterization ......................................................................... 135
3.5.4 Voice Means of Characterization.......................................................................... 137
4 |Page
3.6 Hearing And Perception ........................................................................................... 138
3.6.1 Hearing .................................................................................................................. 138
3.6.2 Voice Sampling. .................................................................................................... 143
3.6.3 Means of Reconstitution ....................................................................................... 143
3.7 Graphics .................................................................................................................... 145
3.7.1 Means of Characterization .................................................................................... 146
3.7.2 GKS....................................................................................................................... 148
3.7.3 PHIGS ................................................................................................................... 151
3.7.4 PostScript .............................................................................................................. 152
3.8 Conclusions .............................................................................................................. 153
4 Multimedia Interfaces ...................................................................................................... 154
4.1 Interface Architectures ............................................................................................. 155
4.1.1 Requirements and Specifications .......................................................................... 156
4.1.2 Elements and Alternatives .................................................................................... 156
4.1.3 Performance Issues ............................................................................................... 157
4.2 Presentation Interfaces .............................................................................................. 157
4.2.1 X Windows ........................................................................................................... 158
4.3 Graphics User Interface (GUI) ................................................................................. 160
4.4 MM Services............................................................................................................. 161
4.4.1 State Machine Analysis and Petri Nets ................................................................. 162
4.4.2 Performance Analysis ........................................................................................... 173
4.5 MM Source Modeling .............................................................................................. 173
4.5.1 Model Elements .................................................................................................... 173
4.6 Conclusions .............................................................................................................. 177
5 Multimedia Storage .......................................................................................................... 179
5.1 Storage Architectures ............................................................................................... 179
5.2 Storage Alternatives ................................................................................................. 180
5.3 File Formats .............................................................................................................. 180
5.4 Multimedia File Retrieval Alternatives .................................................................... 181
5.4.1 Performance Measures .......................................................................................... 181
5.4.2 Access Optimization ............................................................................................. 181
5 |Page
5.5 Conclusions .............................................................................................................. 181
6 Networks and Multimedia ................................................................................................ 183
6.1 Introduction .............................................................................................................. 183
6.2 Architectures ............................................................................................................. 187
6.2.1 Elements ................................................................................................................ 192
6.2.2 Architectural Alternatives ..................................................................................... 198
6.2.3 Impact of Technology on Architecture ................................................................. 203
6.2.4 Architecture versus Infrastructure......................................................................... 206
6.3 Technological Factors............................................................................................... 207
6.3.1 Transport Capabilities ........................................................................................... 208
6.3.2 Interconnection ..................................................................................................... 209
6.3.3 End User Interfaces ............................................................................................... 212
6.4 Market Environment ................................................................................................. 215
6.4.1 Market Players ...................................................................................................... 217
6.4.2 Market Drivers ...................................................................................................... 219
6.4.3 User Value Chain .................................................................................................. 221
6.5 Government Networks .............................................................................................. 223
6.5.1 Structure ................................................................................................................ 224
6.5.2 Competitive Environment ..................................................................................... 224
6.5.3 Optimization Criteria ............................................................................................ 224
6.5.4 Evolutionary Constraints ...................................................................................... 224
6.6 Public Switched Network ......................................................................................... 224
6.6.1 Structure ................................................................................................................ 225
6.7 CATV Networks ....................................................................................................... 228
6.7.1 Structure ................................................................................................................ 229
6 |Page
6.8 Private Networks ...................................................................................................... 232
6.8.1 Structure ................................................................................................................ 233
6.9 Customer Networks .................................................................................................. 236
6.9.1 Structure ................................................................................................................ 236
6.10 Observations ............................................................................................................. 240
6.11 CATV Market Dynamics.......................................................................................... 242
7 Communications Environment ........................................................................................ 246
7.1 Requirements and Architecture ................................................................................ 246
7.2 MM Communications Environment ......................................................................... 247
7.3 Performance Issues ................................................................................................... 248
7.4 Sizing Issues ............................................................................................................. 249
7.5 Architectures ............................................................................................................. 249
7.6 Layers and Standards ................................................................................................ 252
7.6.1 Layer 7: Applications:........................................................................................... 254
7.6.2 Layer 6 Presentation: ............................................................................................ 254
7.6.3 Layer 5 Session: .................................................................................................... 254
7.6.4 Layer 4 Transport:................................................................................................. 255
7.6.5 Layer 3 Network: .................................................................................................. 255
7.6.6 Layer 2 Data Link: ................................................................................................ 255
7.6.7 Layer 1 Physical:................................................................................................... 256
7.6.8 Applications Layer ................................................................................................ 260
7.7 Functions .................................................................................................................. 260
7.7.1 Applications .......................................................................................................... 260
7.7.2 Presentation Layer ................................................................................................ 260
7.7.3 Session Layer ........................................................................................................ 262
7 |Page
7.7.4 Transport Layer ..................................................................................................... 265
7.7.5 Network Layer ...................................................................................................... 267
7.7.6 Data Link Layer .................................................................................................... 268
7.7.7 Physical Layer ....................................................................................................... 268
7.8 Broadband Alternatives ............................................................................................ 268
7.8.1 FDDI ..................................................................................................................... 269
7.8.2 MDS ...................................................................................................................... 271
7.8.3 ATM/BBISDN ...................................................................................................... 273
7.9 Network Management .............................................................................................. 273
7.9.1 Functions ............................................................................................................... 274
7.9.2 Architecture........................................................................................................... 275
7.10 Network Performance and Sizing ............................................................................. 278
8 Session Management......................................................................................................... 279
8.1 Multimedia Data Objects .......................................................................................... 281
8.2 Session Layer Functions ........................................................................................... 285
8.3 Dialogue Management .............................................................................................. 286
8.4 Activity Management ............................................................................................... 289
8.5 Synchronization Management .................................................................................. 290
8.6 Event Management ................................................................................................... 297
8.7 Conclusions .............................................................................................................. 298
9 Distributed Databases ....................................................................................................... 299
9.1 Environment Factors................................................................................................. 299
9.1.1 Network Requirements ......................................................................................... 302
9.1.2 Session Requirements ........................................................................................... 302
9.2 Databases .................................................................................................................. 303
9.2.1 Database Structures ............................................................................................... 304
9.2.2 Data Base Access .................................................................................................. 310
9.2.3 Distributed Databases ........................................................................................... 310
9.2.4 Distributed Database Issues .................................................................................. 313
9.2.5 Multimedia Databases ........................................................................................... 319
9.2.6 Physical Access ..................................................................................................... 319
8 |Page
9.3 Conclusions .............................................................................................................. 321
10 Distributed Operating Systems ........................................................................................ 323
10.1 Operating Systems .................................................................................................... 323
10.1.1 Local Operating System (LOS) ............................................................................ 329
10.1.2 Network Operating System (NOS) ....................................................................... 330
10.1.3 Distributed Operating System (DOS) ................................................................... 331
10.1.4 Media Distributed Operating System (MEDOS): ................................................. 331
10.2 OS Details ................................................................................................................. 333
10.2.1 Specifics ................................................................................................................ 336
10.2.2 Process Management ............................................................................................ 338
10.2.3 Local System Performance ................................................................................... 338
10.2.4 Distributed Systems Design Factors ..................................................................... 339
10.2.5 Device Management ............................................................................................. 339
10.2.6 I/O Management ................................................................................................... 340
10.2.7 Memory Management ........................................................................................... 340
10.3 Distributed Processors .............................................................................................. 341
10.4 Distributed Processes ................................................................................................ 341
10.5 Conclusions .............................................................................................................. 341
11 Conclusions ........................................................................................................................ 343
11.1 Key Issues ................................................................................................................. 343
11.2 Summary................................................................................................................... 345
12 References .......................................................................................................................... 346
13 Index ................................................................................................................................... 359
9 |Page
PREFACE (2021)
This is an edited version of the 1989 class notes on Multimedia Communications which I taught
at MIT. I have tried to keep all the original material and eliminated some material which is so out
of date that it has little use. However, in view of the slow growth of multimedia communications
over this some thirty plus years, I felt an edited and useful version may be of interest. This is
NOT a view from the current time but an edit of what was known in 1989. It is worth looking at
this and comparing what has been accomplished and what is left undone. The basic and
fundamental issues and ideas have not changed. The technology has advanced but the integration
of that technology into a true and working multimedia communication environment has not.
The current virus pandemic has placed an emphasis on the need for true multimedia
communications as presented in this old notes. Two areas are of most import are education and
medicine. Students are left with useless laptops trying to participate in classes using the limited
abilities of a Zoom or some other primitive form of communications. In medicine we have
telemedicine that uses the same limited technology and, in my opinion, does less than what I was
doing at Harvard in 1986-1990. Instead, the current technology implementation has, in my
opinion, wasted decades on such things as Apps while leaving the two areas just mentioned a
wasteland.
There is a philosophy of multimedia communications, a philosophy as to what we are trying to

do as humans in interacting with one another human or group of humans and even our
environment. Multimedia communications is displaced human interactions. It tries to erase the
sense of displacement; it tries to bridge the gap of technology qua technology. None of the
systems currently available have even tried to understand this human interaction challenge. One
cannot accomplish this with a screen on a smart phone. A laptop shining in one's eyeglasses
creates an ethereal image of unreality. Poor speakers and microphones present an almost
comedic interaction, and of course the background is more like a reality home show than a true
normal interaction. Finally, the technology itself gets in the way of interacting.
In 1989 I tried to understand and remedy such issues with a philosophy of multimedia
communications. Unfortunately, most technologists who design and deploy today's systems are
devoid of any understanding of human interactions.
Multimedia Communications must be "displaced human interactions". It is a simple concept but

complex to execute. It must be a seamless and as Heidegger noted "ready at hand" and a simple
"thing" that disappears when used, making the users act and communicate as if the distance was
not there. This is a highly complex challenge. It requires understanding human interaction,
human linguistics and the very nature of humanity at the highest level.
10 | P a g e
This current viral pandemic has demonstrated how poorly we have achieved this goal. Just the
thought of using a smart phone to take university classes, or even primary school classes is an
insane idea. This is the App world of Silicon Valley, a short term monetizable artifact that has set
humanity back, and not allowed us to move forward. How can a Second-Grade teacher interact
with her students? Clearly not with what we have now. How can a father read to his daughter,
not as we have now? How can a Physician interact with a cancer patient unable to come to the
clinic, not as we have now?
Thus, the issuance of this slightly modified revision I believe has value and merit.
I want to also provide a belated thanks to Professor Muriel Medard (MIT) who was my Teaching
Assistant during this first course. She did a magnificent effort in assembling key references and
sources and in managing the class structure. It is to her and my many other students to whom I
dedicate this revision.
Terrence P. McGarty
Thornton, NH January 2021
11 | P a g e
PREFACE (1989)
This text deals with the many issues that relate to the field of multimedia communications.
Multimedia communications is a new field that concerns the creation, manipulation, transmission
and processing of complex images, text, graphics, voice and other multimedia entities. Unlike
the fields of computer communications or even voice communications, the multimedia
environment presents a wide variety of new and complex challenges. The very nature of the
message, complex images as well as voice, and the more complex issues of human interfaces,
make this a challenging field.
This text develops the field by first providing a structure to the multimedia environment and then
developing a theory for its processing, transmission and manipulation. The field of multimedia
communications includes the issues of human interfaces, characterizations of complex images,
the abstractions of distributed data base and distributed operating systems as well the interaction
of multi users in a multimedia environment.
This text grew out of the work that the author has been involved in over a ten year period. This
time period included involvement in such diverse areas as CATV, personal computers, image
processing, broadband communications and ultimately multimedia systems. It had become quite
clear over this period that multimedia communications was becoming a technology unto itself,
addressing many issues that combined the interests of many existing fields but also asking
questions that had not yet been posed. The key to understanding multimedia communications is
not to understand how computers communicate but how humans communicate. Thus the
experience that the author has developed in the context of a multitude of end user applications
has played a critical role in addressing the issues that led to the development of this book.
This book has been written for the use of students who are studying communications, computer
systems or media applications. It is also useful for those practicing in the field and what are
developing new and innovative multimedia systems. It is also hoped that it can be used to inspire
many of those who have entrepreneurial tendencies to implement many of the theories discussed
in this book.
The author would like to thank many of the people who have helped him in the development of
this text at NYNEX and MIT.
Terrence P. McGarty Waterville Valley, NH September, 1989
12 | P a g e
1 INTRODUCTION
Communications has evolved from the simplest form of telegraphy to voice and video and into
the efforts associated with computer communications. The challenge of communications is now
to deal with the ability to use multiple media simultaneously. For example, if we look at the way
creative graphics designer's function, we see that there is high interaction between the images,
the spoken word, the use of facial remarks and most importantly the movement of the hands. The
ability to fully communicate the creative process is the essential element of the multimedia
communication design.
1.1 DEFINITION OF MULTIMEDIA
Media, for the purpose of this book, represents any form of storage or presentation means that
allows for the transmission of information from one user to another. Information in its more
classic sense is basically the reduction of ambiguity or entropy in the understanding of some
concept. Thus the use of a picture, a way of expressing a thousand words, is an alternative means
of providing information through a specific form of media. Images, speech, text, graphics are all
forms of media. The simultaneous use of all of these forms of media in a session dedicated to
information transfer is the essence of multimedia communications.
The history of computer communications is based upon the less than complex transmission of-
information by means of simple binary messages and increasing the complexity to that of text
and simple graphics. The introduction of more complex image formats is increasing and it is the
use of these formats that dramatically changes the way we communicate in a multimedia fashion.
In a simple text communications format, whether it be telex of even high speed computer
networking, the user is forced to follow the standards of the computer network. The messages are
packetized, sending a single message, a single letter or at most a single set of words at a time.
The introductions of graphics packages were really a way to attempt to provide images within
the constraints of the computer network. True image communications, encompassing the full
interactivity of a multimedia information exchange has been severely limited.
A true multimedia multiuser information exchange occurs in a typical ad copy approval session
between the client, the ad agency, the publisher and even the pre-press house. The
communications involves the use of the images, the use of the hands and voice and the
interaction of eyes, body language, charts, graphs, numbers, and finally even the printed words.
Multimedia communications attempts to develop the theory and structure associated with such
information transfer transactions between sets of individuals. Computer communications has
13 | P a g e
typically focused on information transmission and transaction between computers and
infrequently tolerated the interaction with humans, but always on their terms.
At the heart of multimedia communications is the interpretation that this new media-can bring to
bear on our very relationship to what we perceive as knowledge. Marshall McLuhan, the
academic who was made famous by his book "The Medium is the Message" is quoted by Peter
Drucker at the time he was defending his own doctoral thesis. At that time McLuhan was
discussing how Gutenberg had dramatically changed the character of the middle ages with the
introduction of the printing press. As he developed the theses to his academic reviewers, they all
shook their heads in agreement, that is until he reached his final conclusion. That conclusion,
McLuhan stated was that the change in presentation of information changed not only how
knowledge was transferred, but more importantly. WHAT WAS KNOWLEDGE. The medium
of transfer of information actively altered what was information.
We frequently do not readily understand what McLuhan was saying and look at knowledge on a
human time scale. If we were to go back 2,000 years we would see that knowledge was limited
to what could be memorized or at most what was kept on the fragile documents in the few
libraries such as the one at Alexandria. If we look at the knowledge of those times we find a
strong oral tradition and the result was the limitation of ideas to those that had a simple oral
rendering. If we look at Newton's works on gravity we find the vestiges of that oral tradition. The
presentations in written form still follow the oral form with limited equations and even fewer
limited diagrams. In current books on gravitation, there is a plethora of equations and the
diagrams are complex and much more enlightening. As we step even further, the use of super
computers can now allow these theories to be displayed on high-resolution displays in dynamic
form. The knowledge of such thing as fluid flow now is viewed not just as equations, graphs and
figures, but as the flow of simulated fluids through simulate boundaries.
As we move towards the current days with the use of film and video, we find that knowledge is
viewed as what is on film. The anecdotes, whether they are true or not, of recent presidents
believing reports only if they are on film rather than written, show again how the change in
available media change what is knowledge. Knowledge is the image or vision of the camera and
its ability to position certain ideas in the eyes of the viewer.
As we move forward in the use of new media, especially those that allow for multimedia and
multiuser interactions, we further build upon McLuhan's concept that changing media changes
what is knowledge. The book as a learning medium will change. The book generally is a static
medium that is linear in form, progressing from chapter to chapter. The book is built up by the
author to present a sequence of concepts to the user. The book is non interactive, it does not
allow for the asking of questions and in response provide a new reordering.
14 | P a g e
As Nathan Felde has stated, education is a low bandwidth process, entertainment is a high
bandwidth process. Education is a questioning nonlinear an iterative process that frequently uses
many media. Entertainment is a nonquestioning linear process that lacks significant interactivity.
We must recognize that the multimedia user frequently is trying to be educated, that is transfer
and absorb information, not be passively a recipient of high bandwidth entertainment.
1.1.1 The Image
We shall expand the concept of an image to mean the embodiment of any media based
information bearing element. For example, an image may be a picture, a segment of a voice
conversation, a piece of graphics display, or even a segment of binary data. In the general sense,
if we let M represent any such image, then we define the image in terms of is extent, E(x,y,z,t),
its information content, I, and its location L(x,y,z,t).
The image will be at the heart of the multimedia communications environment. It is the
generalization of the basic element that we shall consider at part of the transfer of information. In
a voice only environment, the corresponding element is the conversation, the concatenation of
the words, inflections, pauses that make up the communications from one individual to another.
In a computer communications world the corresponding element is the data file and the
transactions that relate to that file. As we expand to the multimedia environment we extend
beyond that to the more complex interaction of user with information in many forms.
1.1.2 The User
In a multimedia communications system, the user is generally not the large computer or other
information processor but is the human. This dramatically reorients the focus on what is to be
communicated and how it is to be communicated. The users objective in a multimedia
communications environment is to use the information and to develop and understanding on how
the many parts of the information can be used—in achieving "the desired" goal.
1.1.3 Interaction
The interaction allowed in a multimedia environment is dramatically different than that in most
other communications systems. As we noted before, the standard form of communications is
highly structured and is linear in fashion. Computer communications is based upon the need for
well-established and agreed to protocols that permit the users to interact in a controlled fashion.
The users in the computer world are generally other computers or in some cases humans who
must adhere to the computers well structured dialogue. In contrast, the interactions in a
multimedia environment are highly unstructured, significantly nonlinear in their form and are
15 | P a g e
driven by the more creative side of the human user. Images are complex representations of
knowledge elements and the ultimate information is a construct of the combination of the
images, the user and the interactions that are developed.
1.2 CURRENT SYSTEMS
There are several simple but representative multimedia system that have been developed in the
past few years. The first contain work stations or computers that allow for the use of many types
of media, integrating graphics, still images and other types of images—into the screen. The
introduction of a
window environment allows for the use of many of these media on a standalone work station.
The inclusion of a networking capability further enhances the ability of these work stations to
share the resources. For example, a file server is a memory storage device that allows for shared
memory to be used by many users on a local area network. The server may have any image type
stored on the system and one or several users may window the image onto their customized
screen format.
Figure: Sample Current Systems
16 | P a g e
Figure: Sample Current Multimedia Systems - the DARPA Experimental Multimedia System.
17 | P a g e
1.2.1 Computers
The multimedia work station or computer has been developing over the past few years and a
typical design is that of the Apple MAC II system. The MAC II uses a simple window structure
with a high resolution bit mapped scree for display. We shall see that this type of work station
architecture is typical of many of the first generation multimedia work stations.
Figure: Sample Multimedia Computers
18 | P a g e
Figure: Action Flow
1.2.2 Presentation Formats
The first generation multimedia systems allow for a wide set of presentation formats. The
systems will support graphics, text, still images, video of various forms, digitized displays, and
integrated voice with the display. Also some systems under development allow for the
integration of voice and image as an integrated whole. Some systems allow for the integration
and manipulation of various sound elements.
19 | P a g e
Many of the efforts in the MIT Visual Language Workshop, under Prof. Muriel Cooper, show
the capability of combining the visual with sound of many forms.
In a multimedia environment, it is possible to interact with all of the human senses and more
importantly to play one off against another. We can exchange sound for intensity and color for
the pitch of speech. In the current first generation systems, there is considerable effort using the
concept of false colors to bring out new information in visual data that was no there in a simple
viewing. A typical example is what can be done using extensive color imaging in the medical
area. Currently all X-rays, CAT scans and MRI scans are in gray scale images. A gray scale
image quite simply is a system that put at each picture element (pixel) a dot whose intensity can
vary from pure black to pure white and has many shades in between. For example the display at
each pixel may have 10 bits and thus there could be 1024 different shades of gray. Depending on
the human, the eye could resolve anywhere between 256 to- 4096 shades of gray.
Now it is possible to expand the gray scale imaging of human tissue by adding color, wherein the
color corresponds to tissue type and its intensity to type density. Thus it is conceivable that an
MRI scan could look at the paranasal sinuses and determine not only the different masses and
size of tissues but could also identify particular tissue types. The discriminants to do that are
available today in the magnetic resonance imaging data.
Figure: Sample Integrated Presentation Formats (Muriel Cooper's)
1.2.3 Interaction Formats
20 | P a g e
The end user can interact with the images in many forms. The standard first generation
interaction mechanisms include the keyboard, the pen and pallet, the touch screen, and even a
voice activated display. Some extreme forms of interaction have included the are/hand point and
speak methods developed in the MIT media Lab. The latter types represents some of the first
attempts to make the interaction more in line with the activities of the human user. Remembering
that multimedia communications attempts to match the end users flexible forms of inputs, the
need exists for many types of advanced interaction formats. Attempts have been made to provide
for speech response formats that allow for hands off interaction. The problem generally with
these types of devices is that they displace the problem of input to a higher human level of
making the user coordinate two separate reflexes, sight and sound.
Considerable effort is still required to develop interaction devices and systems to more
effectively match the human user. The classic anecdote is the scene in the movie Star Trek,
where the chief engineer Scotty, sent back sever hundred years in time, asks for a computer and
is shown an Apple MAC. He picks up the mouse and speaks into it and says, "Computer". He is
then told that that is a mouse and he must use it to point and the key board to enter the
commands. He comments, "How quaint". Yet it is this interaction element that will be the main
diver of the McLuhan change of media and knowledge.
The interaction with the current first generation systems focuses on the interaction of the human
user with" the stand alone work station. We can envision extending this by having a network of
resources and other users and consider the interaction problems in this multimedia multiuser
communications environment. We shall, in the remainder of this book, develop the concepts
necessary to show how this more complex system of interaction may evolve.
1.3 THE PARADIGM
The development of a theory for multimedia communications requires the development of a

paradigm or world view of the multimedia environment. We have defined the concept of an
image and we must now combine together several key concepts that will be sued over again in
the book as we develop the multimedia communications environment. All multimedia
communications relate to images. Yet those images ultimately relate to the interaction of human
users and their mutual manipulation of those images.
We shall now define the key elements of the multimedia paradigm and show how those elements
add to a working model for multimedia multiuser communications.
Figure: The Multimedia Paradigm
21 | P a g e
1.3.1 Individual User
The human that interacts with information is termed the individual user, IU. In any
communications interaction we can envision one or more individual users communicating. The
individual user in multimedia communications becomes a key element in the ' communications
channel. The individual has an identity that is separate and apart from the other elements that are
frequently the key elements in a computer based communications example. Even in the standard
world of voice communications there is no recognition of the individual directly. The telephone
directory gives only the name of the resident responsible for the payment of the bill. In
22 | P a g e
multimedia communications, we recognize individuals separately and apart from the other
elements in the communications system.
1.3.2 Physical User
The first element in the model is that of the physical user. The physical user is a physical device
that is used by an individual for the purpose of communicating with others. Clearly, we can
envision that the individual may have access to several physical devices and in turn a single
physical device may relate to several individual users.
The physical user, P, may be one of several physical devices that are attached to the network.
These devices may be terminals, file servers, imaging devices or there elements. They are unique
and are specifically located at a specific location.
Virtual User
The virtual user, V, is a concept that becomes an important element in the multimedia
environment. Any physical user may open one or more virtual users on the network. To some
degree, the concept of a window is analogous to that of a virtual user. Any physical user may
have several virtual users connected to the single physical device.
Application
An application is a specific program that is operative within a physical user environment. For
example, in a CAT scan environment, the physical user is the CAT scan system, the virtual user
is the specific allocation of resources to a set of scans and the applications may be a combination
of both two and three dimensional CAT scan analyses. The virtual user window can open up for
the analysis of a specific patient. Thus the identification of a virtual user in this case is with the
patient, and within that virtual user is allocated two applications programs that process the CAT
scan data and another application program that is used for the maintenance of the hospital
records.
Image
The image is the physical embodiment of the multimedia subject to be used in the discussions
between users in t:he session. Tor example the image could be a picture scanned from a video
data base, a graphics image generated from a raster or geometric data format, or even a voice
message that is interactively generated between the two users.
Object
23 | P a g e
The object is the first step towards abstraction of the image. Whereas the image is the totality of
the multimedia message, the object is a bounded portion of the image that can be user definable.
Thus if the image is considered to be the full landscape of a photograph, the object may be the
part of the landscape that is near a stream or mountain. Objects present localization of images. A
further example is that in the general sense the image may be a total conversation, whereas the
object may be a portion of that conversation.
Element
The element is the part of an object that can be abstracted to a specific definable abstraction. For
example in an image of an X-Ray, there can be a specific object called the spine, and a specific
element called the fifth cervical vertebra. The ability to now abstract the fifth cervical vertebra
and use this in comparison in MRI, CAT and radiographic studies is the key issue associated
with elements.
Figure : Multimedia Elementals
24 | P a g e
Figure: Multimedia Elementals
25 | P a g e
1.3.3 SubSession
A sub-session is a connection between two virtual users. It is the fundamental building block for
multimedia multi user communications. What is important
1.3.4 Session
The session is the key concept in the multimedia communications environment. A session is
quite simple a collection of one or more sessions. In a session, a set of virtual users, and their
associated individuals and physical users, are linked together into a common communications
process. In this book we shall be developing all of the structure necessary to develop the session
and to support the session as it evolve in time. A session may last an indefinite period of time
and it can survive the coming and going of any of its constituent sub-sessions.
All interaction occurs within the context of a session. In addition the subsessions allow for the
adding of any set of virtual users, each of whom may have their own objects, elements or even
applications.
Figure: Development of Sessions and Subsessions
26 | P a g e
1.4 ELEMENTS OF MULTIMEDIA SYSTEMS
We have discussed the structure of the current first generation multimedia systems and have
discussed the way in which they process data on the work station, display the images and
integrate them and provide for interaction with the user. The first generation systems are
generally limited in what they can do in these three areas. In this section we shall develop the
basic elements of the multimedia communications systems and attempt to anticipate what is
possible in the second generation of such systems.
Figure: Multimedia Layering and Interfaces

27 | P a g e
28 | P a g e
Figure: Multimedia Layering and Interfaces - Interface between two ISO Layers.
Link Network Transport Session Link

Protocol Protocol Protocol Protocol Protocol Control Information and
Data in Transfer Syntax Check
Control Control Control Control
Sequence
Information Information Information Information
29 | P a g e
Figure: Multimedia Layering and Interfaces - Layering in the Data Packets.
1.4.1 Display
The display for the first generation systems has generally been a bit mapped display whose size
has been limited to about 1000 by 1000 dots. This yields about a million pixels in a display. The
pixel may be up to 12 bits deep, so that a single display has 12 million bits. In current production
as the first steps towards a second generation display are 2000 by 2000 units of 4 million pixels,
each of 24 bits per pixel. This yields 100 million bits per screen. This approaches the display
capability of a 35 mm slide.
30 | P a g e
These second generation displays now must have significant buffer storage to keep the image
and must have the bandwidth at the display driver to keep the image refreshed at the rate on 30
times a second. This implies a transfer rate of 3 Gbps.
Figure: Display Architecture - A Simplified View of the Vector General 3400 Display Architecture.
1.4.2 Applications Software
The applications software must handle the processing necessary for the display processing as
well as for the support of the end user interaction. It is a layered approach to the development of
the application software, the lowest layer interfacing with the operating system of the multimedia
environment and at the highest layer, and end user interaction capability. The layers include the
following:
User Interface Layer: This outer layer allows for support and integration of the end user into the
system.
Applications Support Layer: This layer provides for the interface to the User layer and utilization
of the local work station for support of local processing.
31 | P a g e
Data Management Layer: This layer provides for the support of the local data files and
management of the images into the display. Such capability now in the X Window format can be
found resident in this layer.
1. Operating System
2. Control
3. Supervisor Programs
4. I/O Programs (IGES, COM)
5. Communications Programs
6. Processing
7. Utility Programs
8. Language Translator
9. Information Base
10. Management
11. Application
12. Pagemaker, Videoworks, MacDraw n, Quark Express, (Core, GKS or PHIGS standards for
graphics) Legend: - GKS : Graphical Kernel System PHIGS : Programmer's Hierarchical
Interactive Graphics Standards.
Figure: Software Architecture - Software Architecture for a Multimedia Messaging System.
32 | P a g e
1.4.3 1.4.3 Services
The services portion of the canonical design allow for the support of the multiuser environment
and are the key to enabling the establishment of the session. Typical services may include:
Session Establishment: This service allows for the generation and support of a session between
any of the users. A single user may establish a session between themselves and any other virtual
user available on the network.
Mail
File
33 | P a g e
Directory
The services can be implemented in a layered architecture just as we had done with the
applications layer. It consists of two layers, the shell and the kernel. They shall allow for the
interaction between the individual terminals and the services and the kernel supports the
interaction between the individual terminals and the network. The key to successful multimedia
services is the ability to work in a fully distributed environment. We shall develop the distributed
environment in latter chapters and shall also present a more detailed structure to these two layers
of the services.
1.4.4 1.4.4 Network Software
The network that interconnects the end users must do so in support of all of the layered areas
discussed above. The network must support the applications areas and furthermore support the
services layers. The major difference between multimedia communications and computer
communications is that often the computer needs to send the messages in what is called a
connectionless format. Packet or datagram are sent as little letters with all the information to get
from point A to B. In the multimedia environment, the need for the use of sessions requires that
the sessions be supported by a connection based network service. In addition the network must
be capable of providing large amount of bandwidth on demand to the user. Consider the case of
the second generation high resolution display with 100 M bits of data. To transmit one image
requires a 100 Mbps channel from any point to any one of a set of other point. The response time
of less than one second is critical to the success of any multimedia system.
This architecture can be either distributed or centralized in nature and the network software must
be intelligent enough to support the session based connection path. We shall be developing this
concept in further detail in the remainder of the book.
1.4.5 1.4.5 Storage
Multimedia communications requires large amounts of complex image information for use by
many users all at the same time. No matter how the cost of memory decreases, the key bit of
information will not be sent everywhere at the time it is generated. Thus the network must
provide access to memory that is distributed and that also can be accessed in extremely short
periods of time. Consider again the example of the high resolution display that requires 100 M
bits of image. The transport speed may be 100Mbps to lGbps but the bottleneck may be in the
finding of the imaging, accessing it and loading it onto the network.
1.5 KEY PROBLEMS
34 | P a g e
One of the world's greatest philosophers, Ludwig Wittgenstein, once stated that the essence of
true knowledge of any field was the ability to pose the correct set of questions, whose answers
are simply stated. Thus with multimedia communications, the development of a new body of
understanding is best based upon the positing of the proper set of questions and these questions
are in turn based upon a set of key problems that we face in attempting to communicate in this
new environment. The following problems and the ensuing questions represent the body of the
text.
1.5.1 Problem 1: Characterization
As we have developed in the early part of this chapter, the essence of multimedia
communications depends on the concept of the image. The image may vary in form from a high
resolution picture to a speech conversation. The first problem is the characterization of the
physical image into an electrical form. For example, a speech conversation that varies over a
finite period of time may be digitized in a standard form by sampling at the rate of once every
125 msec and using eight bits per sample. This will yield the standard 64 Kbps voice sample of
speech.
As a second example of the characterizations problem is that applied to high resolution pictures.
Again we can consider sampling the image in two dimensions and in storing the image in this
highly sample digital form. The problem is again how well to - sample to retain-the integrity of
the basic imager
Another element of the characterization problem relates to how to characterize an image so that
it may be manipulated and not just stored. Consider the following example.
Example: In the medical field, there is significant use of both CAT (Computer Aided
Tomography) and MRI (Magnetic Resonance Imaging) imaging. CAT scans use X rays to
resolve the different parts of the body tissues based upon the density of the tissue and its ability
to absorb the X radiation. By passing many different X rays through the body the image can be
reconstructed to reproduce the innermost portions of the body. In MRI, the technique uses the
ability of different tissues to resonate when under strong magnetic fields, the resonant
frequencies dependent upon the tissue type. In the MRI case a multi-dimensional Fourier
transform creates images of the human body. Now the physician desires to compare the results of
the two scans to determine the status of the cerebellum, the posterior portion of the brain. The
characterization problem in this case relates to how do we define the cerebellum as an
abstraction in both imaging schemes. We wish to do so not by just comparing bit mapped
displays of the same tissue but to do so by a full element abstraction.
1.5.2 Problem 2: Utilization
35 | P a g e
The utilization problem takes the image characterization and addresses the issue of the human
interface. How does the human specify the specific image, object or element, and in turn how
does the human user interact with the system to allow for the manipulation of the images. The
concepts of the touch screen and the mouse are but simple starts to the development of agents in
this area. The general concept of an agent, that is an entity that can abstract both and element and
an action are key to understanding the issue of utilization.
Transmission of the information relates to the problem of taking the users view of the world and
allow for it to be effective in several places at the same time. In contrast to sharing, which we
shall describe in the next section, transmission relates only to the transport for the images in a
session context. The present means of transmission are generally quite limited. They use existing
telecommunications services that in most cases are not designed for high resolution image
transport or for the use of extensive session sharing. The telecommunications transmission
channels allow for rates up to 45 Mbps which limit the use to small image sizes or lower refresh
rates.
For example, consider a 2,000 by 2,000 pixel array using 24 bits per pixel. This amounts to about
100M bits per image. If such an image were to be used in a full motion video context at 30 times
a sec refresh rate, then we would need 3 Gbps transmit capacity. That is almost 100 times greater
than the current offered tariffs for transmission services. Thus the transmission problem focuses
on two issues, the first is developing higher data rate transmission channels and the second is the
developments of compression techniques for the reduction of the data required to transport the
true information content.
The true transmission dilemma that faces many users is to wait for the greater data rates or to
incorporate the use of often high cost compression technology. The date rates that are now being
implemented are moving to the 500Mbps range for local access and to 1.5 6bps for long distance
transports—The -problems still exist that one cannot share the data links and thus one is
compelled to use the data channels as if they were fully loaded.
1.6 PROCESSOR/ PACKETIZER
The sharing problem relates to how best to take multiple users in different modalities of use and
to allow them to share the communications and end users services in a user friendly fashion. The
essence of sharing is essentially the development of the session concept that we have developed.
As we have stated, the session concept was developed to support the virtual user layer of our
multimedia architecture and allows for the interface and support at that level. The applications
layer interface will be shown to have the session capability already present in what is called the
OSI Layer 5 Session protocols.
36 | P a g e
The complexity of the sharing problem or that of VU Sessioning relates to the need for
connecting diverse end user devices together with diverse end user interfaces. If there were
standards that existed for doing so the problem would be generally easy. Such standards exist for
the applications to applications interface but fail to exist for the VU to VU interfaces. We spend
a great deal of time in this book focusing on the issue of sharing.
Example: Consider the medical application again wherein the user has to work in a complex
environment of CAT, MRI and Nuclear medicine scans.
The applications to multi user systems is characterized in the sharing problem but the multimedia
multi user system problem is best characterized by the mixing problem. The mixing problem is
the one that occurs when different forms of images and media are to be shared and manipulated
by multiple users at the same time. These users may have different types of equipment, have
their images on different types of databases and different interfaces.
Figure: Mixing Concept in Printing Environment
Figure : Mixing Concept in Printing Environment -Mixing in Multimedia Documents Using an Area
Control Mixing Type.
37 | P a g e
1.7 OUTLINE OF BOOK
The book follows the discussion on the key problems that frame the multimedia multi user
communications area. In this book we spend a great deal of time discussing the characteristics of
many forms of multimedia images. Specifically we discuss the way in which users of still
images, graphics and video, for example characterize their media. This is important since when
we take these existing media that are in a physical and non-electronic form and turn them into
electronic, we must do so non-electronic as to preserve their original frame of reference. For
example, if we develop a system to display radiological studies from file to a digitized form, then
we must preserve the same image density, intensity and responsivity as does the film radiograph.
38 | P a g e
discusses the many types of multimedia interfaces. These interfaces range from the ways in
which we transform the physical media of the image into an electronic image to the ways in
which we characterize images in abstraction forms. As we have developed in this chapter,
abstraction of images is one of our major concerns. The approach to abstraction may be
structural or connectionist. Our major focus in this text is that of a structural approach, leaving
the connectionist to those other works that cover them more fully.
The issue of storage of multimedia information is contained in . In this case we are concentrating
on multimedia databases that are significantly more complex than those found in normal data
processing environments. The multimedia database problem revolves around the issue of the
scale of the image and its need for integrity and simultaneity. Another element of the
multimedia communications that this chapter considers is that of performance and sizing. These
latter two items are of critical importance in designing and developing any new multimedia
system.
addresses the issues of communications. We focus on the standard approach to communications

by working with the OSI layers. Our approach is different than most others by starting with the
higher layers, viz. applications and presentation and session, and working downward. The issue
here is that instead of looking at the physical communications link as transport and then finding
what is missing, we focus on the session and the essence of the multimedia environment and
works downwards. This approach provides new insight into the communications link. We also
develop an understanding of the systems and issues associated with broadband communications
systems.
The essence of any multimedia system is the development of an overall architecture. The first
few chapters focus on the major problems that we have developed earlier in this chapter.
develops the overall architecture for such systems. In particular we focus on developing a
methodology for architecture development, including the issues we have developed on
characterization, processing, storage and communications.
The book develops the theory of multi user systems, showing how to extend the elements of the
multimedia environment to that of the many user system environment. We spend a great deal of
time discussing the performance and sizing efforts of such system designs. It develops in details
many of the issues associated with the implementation of distributed systems. In the multi user
multimedia environment, we are dealing with a fully distributed environment. In this
environment we must focus on distributed databases and distributed operating system. As we
have discussed the multimedia database we must now focus on how it can be expanded to a
distributed environment. The issues of distributed operating systems allows for a better
understanding on how multi users may now deal with the issues of sessions in a shared
39 | P a g e
synchronous environment. One of the major issues is that of understanding the overall synergies
of a shared network.
Finally it presents the overall conclusions and the directions some of the multimedia system are
taking in today's environment. In particular, we discuss many of the issues that will become
research issues in the years to come.
40 | P a g e
2 MULTIMEDIA COMMUNICATIONS
From time to time, a change occurs in the way people communicate, to cause a shift in thought,
perception, awareness, and in a sense of what is truth and knowledge. It has been argued that
multimedia and multimedia communications is such a shift. This paper addresses that perception
from a perspective that is dramatically different than any other taken. The perspective developed
in this paper is one that is philosophical. It does not talk to technology or business. It does not
refer to companies and equations. The approach is that taken by an engineer who was first
trained as a philosopher. An approach that tries to address first principles from the context of the
perceptions and understandings of the many other thinkers who have addressed the issues of
human knowledge and communications before. We consider the process by starting with Plato
and end with the most current thinking in the areas of epistemology and human understanding.
This is an evolving approach. The schema presented and developed herein are still formative.
They do, however, set a direction and a discipline.
2.1 CURRENT DEFINITIONS OF MULTIMEDIA
Multimedia concepts encompass a wide variety of interpretations. To some, the multimedia

environment consists of one that has many elements of storage, to others the environment entails
the interfacing of various storage devices. In another extreme there is the concept of a
multimedia environment being nothing more than that of a windowed display environment that
uses some limited form of hyper media, that is windows in depth besides just length and width.
We attempt to develop a theory of multimedia communications, that encompasses the elements

of both the multimedia environment as well as the effective utilization of the rich nature of that
environment. Let us begin by understanding the nature first of the human using information in
the context of their own limited environment and then expand it to the case where the human can
share the information with others.
There have been a significant number of definitions of multimedia. Comparing some of these
definitions will be a quick way to determine what multimedia is not. All too often the authors are
reporters who have not struggled with the issues of multimedia or at the other extreme they are
vendors who define multimedia to be what is in their current product line. The first definition is
due to Wright, who describes multimedia in the context of the Global Village. To her,
multimedia is;
" Computer based presentations combining two or more media, such as text, graphics, writing,
and video and audio signals."
41 | P a g e
If we look at her definition we note that she defines media in the context of the type of message
(text, graphics, video, audio) as well as the way it may be stored (video and audio signals). She
does not take into account the complexity of storage, the sophistication of interfaces, the
interaction of the senses, nor the intelligence of the human processor. The mere fact that
multimedia is "Human" directed as compared to "Computer" directed is not an element in her
definition. As a point of fact, Wright goes as far as to define the Human Being as:
"An analog processing and storage device with the bandwidth of 50 bits per second. Human
beings excel at pattern recognition but are notoriously slow at sequential calculations."
One is amazed that the author fails to recognize that the essence of multimedia is to more
effectively match the input/output capability of the human as a species. We humans are
amazingly fast at processing visual information, less fast at aural information and slower at the
use of our other senses. Ms. Wright in her Scientific American presentation totally misses the
essence of this new paradigm.
Probably one of the more insightful presentations of the Human as processor is the book by Max
Delbruck. Delbruck, the Nobel prize winner in Biology, presents the argument that the evolution
of primates, and humans especially is epitomized in the evolution of vision as the primary sense.
Three specific developments allowed for this speciation to create the mind of the human. The
events are;
1. A transition from an olfactory and tactile creature to a visual creature.
2. The displacement of the eyes from lateral to frontal allowing for the perception of three
dimensions.
3. The differentiation of the cone receptors in the retina allowing for color perception.
Starting with these three events, Delbruck demonstrates that the human mind is intimately
connected to these events, and as such is a complex image processor. This is the basis of the
multimedia paradigm. This is what Wright has totally missed in her exposition.
A second presentation of multimedia is done by Jane Morrill. Morrill defines multimedia in the
following even more interesting form.
"Flip charts? Overheads? Isn't this the computer era? Surely, with all the high-speed machines,
CD-ROMs, synthesizers, and image processing capabilities available, there must be something
42 | P a g e
that will convey your message better than flip charts and overheads. Well, there it is. It's called
multimedia, and it marries the best of image, voice, text, and video processing."
Her attempt at a definition is a tautology, in that you know what it is when you see it. Her article
then proceeds to develop multimedia in the context of what vendors of workstations have in
terms of different storage devices. She again does not even acknowledge the issue that the human
is a factor and that multimedia is a blending of the human in the stream of the information flow.
A third paper by Shandle defines multimedia in altogether different terms. Specifically he calls
it:
"...-the delivery of video, audio and other heretofore exotic data types to the desk top -...."
Again there is no mention of the human, why the data is to be transported or even what it is to be
used for.
2.2 PHILOSOPHICAL PILLARS
To fully understand multimedia it is necessary to explore the work of two sets of major thinkers.
The first is Marshall McLuhan and the second is Winograd and Flores. McLuhan, in
Understanding Media, and Winograd and Flores in Understanding Computers and Cognition
have complemented each other in a way in which their convergence of ideas lays the ground
work for "Understanding Multimedia Communications". We shall be relying on these two sets of
authors for a guide through the development of the meaning of multimedia. McLuhan has been
discredited of late because of his simplistic views. We shall argue and shall attempt to show, that
this may be a direct result of the critics, frequently the Pop Press, not understanding how
perceptive McLuhan was in his more academic treatises. McLuhan will be the definer the "bright
line" that results when a paradigm shift occurs in a new medium.
Winograd and Flores are the other set of lights that the author shall rely heavily upon. Unlike the
Pop writers of the above definitions, Winograd and Flores have developed one of the most
seminal works in the areas of computers and computation that have ever been done. These
authors have used access to the most recent philosophical understandings of knowledge and
knowledge processing from a philosophical perspective to develop a philosophy for computing,
especially software development. We shall, in this paper, develop and extend these concepts for
the multimedia area.
Drucker, in his biographical sketches of his contemporaries, remarks on his first encounter with
McLuhan. It was during a presentation that McLuhan was making on the results of his do
doctoral studies. His presentation reflected upon the impact that the printing press has had upon
43 | P a g e
the university system in the late Middle Ages. He contended that the modern university came
into being in the sixteenth century because of printing, which changed not only the method of
instruction but, more importantly, what the university intended to teach. He further contended
that the cultural results of this period had little to do with the Renaissance and was all a direct
result of the printing press.
To quote Drucker, who paraphrased McLuhan;
"Did I hear you right," asked one of the professors in the audience, "that you think that printing
influenced the courses that the university taught and the role of university all together." "No sir,"
said McLuhan, "it did not influence; printing determined both, indeed printing determined what
henceforth was going to be considered knowledge."
Thus this led to McLuhan's famous phrase that the medium is the message. Specifically, as we
developed a new medium for human communications, we dramatically altered the nature of the
information that was transferred and the way in which the human perceived what was "truth" and
what was not. The television generation of the 1960's was an clear example of the impact of
television versus film in portraying the war in Vietnam as compared to the Second World War.
The perception of these two events was determined by the difference of the two media that
displayed them to the pubic masses. Television allowed for a portrayal that molded more closely
to the individual humans impact of the events as compared to films overview of the groups
involvement's. Both media deal with the same senses but they are different enough to have
determined two different outcomes of the wars. This conclusion is a McLuhanesque conclusion
but is consistent with the changes that McLuhan was recounting in the 1960's in his publications.
The important observation that McLuhan makes is not often understood. He really means that the
medium defines what is knowledge. A new medium, as a general construct, will define a new
knowledge base. We all too often define knowledge so obtained with truth. In fact, truth is that
relative reality that we find comfortable to our understanding, and all too often ascribe an
absolute character to it. The essence of this paper will deal with these two issues; knowledge as
defined in the McLuhanesque sense, and truth as a phenomenological expression of that
knowledge. Multimedia communications will alter those definitions and will dramatically change
the way we see, think, and ultimately act. We argue, for example, as with McLuhan, that
television violence, for example, changes what is knowledge, the acceptance of moral norms,
and this change in moral knowledge is reflected in the truths of everyday existence. The
expansion of multimedia communications will take this minor concern many levels higher. Thus
multimedia communications is a technological issue, a philosophical consideration, and
ultimately a moral imperative.
1.3 Multimedia Structures
44 | P a g e
Multimedia consists of three major dimensions; the storage media, the interface media and the
senses. A true multimedia environment provides a full mix of all of these. Let us begin by first
considering the senses. The five senses are; sight, sound, touch, taste, and smell. As humans, we
specialize in the first three senses, and only secondarily allow for use of the last two. Max
Delbruck, the noted biologist and Nobel prize winner has discussed the uniqueness of the human
as one whose dealings with his environment is more visual than olfactory than any other species.
Most animals have highly developed senses of smell and frequently their total social patterns
revolve around them. We as humans revolve around sight first and sound second. Point of fact,
as cultures evolve, we find that the use of the other senses are less and less socially acceptable.
We no longer sniff at each other, nor do we employ the other two senses as other parts of
Animalia do.
The senses element of multimedia provides a basis for the understanding of the limitations and
emphasis of the human as a processor of information. The information is contained in a
combination of sensory inputs. These inputs are all parts of the communications environment. A
discussion of an ad layout for a print publication involves the image, the voice interaction as well
as the human interaction of peripheral nature that ensues during the presentation of the ad copy.
To develop multimedia effectively, it is necessary to evoke all elements of the human response.
The interface element of multimedia communications describes how the human interfaces with
the electronic interlocutor, the electronic medium. Again we all too frequently view the interface
in simple terms as a computer display and nothing else. The current view of multimedia
communications is that of a display terminal that allows for both a windowing environment and a
hyper media environment. It may also extend to including video. As we have just described, the
essence of multimedia is to maximize the input from the senses, and also balance the
communication of all sensory data on the topic to be discussed.
Thus the interfaces must communicate all five senses both from the electronic medium and to the
medium. Visual displays are but a part of that communications. Effective multimedia today must
also integrate voice and sound with some form of tactile interface. A now primitive example of
this is the first set of electronic games wherein the human had visual, aural and tactile interfaces
with the electronic medium.
The final element of a multimedia environment is the storage media. The storage media in some
sense dictates what is effectively communicated and the ability of the human to retrieve and
process the information. The media today has allowed for the storage and retrieval of
information in short times for large amounts.
45 | P a g e
Here the author defines Multimedia Communications in a broad context. "The term multimedia
is used in many different contexts and is greatly misunderstood. In the current popular press it is
viewed as merely a set of mechanisms for the storage of different data types on a local basis and
their display to a single user. In the extended view of multimedia communications it is
understood as an environment for the sharing of information in various forms, video, image,
voice, text, records, etc., that are stored in different locations. Diagnostic imaging places
significant challenges upon the diagnostician as well as the attending physician and multimedia
communications can assist in meeting this challenge. There is an increasing need for the set of
physicians to act in concert in combining their talents for the best delivery of service to the
patient.
Multimedia communications is characterized by the following factors:
1. Multi-Sensory: It uses several of the human senses in transferring, processing, and

creating information.
2. Multi-User: It interconnects several users of the information into a conversational mode

and allows a dialog based on a fully interconnected set of media.
3. Displaceable: It allows for the establishment of communications and information transfer

that is displaced in both space and time from the source.
4. Interactive: It permits a real time interaction between any of the users of the medium,
whether the users be human or databases or applications software.
The multimedia environment is one that is user centered and is designed to meet the users' needs
in interfacing with complex images and in conveying information from one location to another.
Multimedia is not just a description of how the data is stored, it is, more importantly, the
description of a philosophy of human interaction with complex data elements in a multi-sensory
fashion."
We must ask the question of what does the human want to do with multimedia that the human
does not either want of is able to do with the classic single media information sources. There are
several processes that are necessary;
1. Define: The user desires first to define a multimedia object. This object is in sharp
contrast to a normal data object that is typically a structured and bounded alphanumeric data
element, convertible into a digital representation. A multimedia object is the concatenation of
video, voice, text, and other sensory representations of the event at hand. A multimedia event is
the analog of a data object. The data object is the representation of a definable and measurable
46 | P a g e
term used in common communication, such as the word, NAME. In contrast the multimedia
event is a collection of multimedia presentations that have a temporal and spatial extent to them.
That is a multimedia even is a set of voice segments, a set of video elements, and a set of text
frames.
2. To define a multimedia event means to concatenate in a rational form the set of disparate
multimedia elements into a connected event. The process of connection is complex but it goes to
the heart of human communications and understanding.
3. Query: The query in a data object case is a way to do one of two simple tasks. A data
object may be either selected or enumerated. The selection process is based upon the ability to
take a data object and recognize that it has a unique representation in some stable denumerable
set. For example we can use the alphabet or a binary representation for any data element. Once
we select a data object we can then enumerate all of the objects that meet a certain criteria, by
again matching and now counting. Thus we can answer the question of how many patients over
forty have high blood pressure. We first use the select process on high blood pressure and then
the enumeration on patients. In a multimedia environment, we are now posing much more
complex queries.
4. Store: This means that we must store complex multimedia objects, composed of video,
voice, image, pointer movement, text etc. that may reside on different storage devices at different
locations. We must be able to retrieve them in the same order and timing that they were stored
and do so in a minimum time.
5. Process: We must be able to process multimedia objects, to alter, enhance, combine them.
We must perform the processing in a fully distributed fashion, using the resources from multiple
processors.
6. Display: The display of the multimedia objects includes not only the display of the image
or video but the "display" of the voice and other sensory elements of the multimedia object.
7. Communicate: Communications means the development of a conversational mode.

Conversation is key to communicating in a multimedia environment. Thus, we must not
reproduce a communications environment that is attuned for the computer but one that is
matched to the human user. The essence is the ability to effectively share the multimedia objects
in a dialog fashion, interactive and interpretive.
These multimedia processes are to be done in a fashion that is transparent to the user. They must
also be done in a fashion that is resonant with the way the users currently performs the tasks.
47 | P a g e
There are several dimensions that can be used to characterize the extent of the multimedia
environment. These dimensions are;
1. Time and Duration: This dimension shows the amount of simultaneity that the medium
allows both for a single user as well as for a collection of users. Further a dimension of durability
to the environment is essential as the complexity of a multimedia object requires that time pass
until it has its full representation. Thus unlike a mono-media object that can be represented to a
single user in a fixed period, the interlining of media and users requires a sustainability of the
environment.
2. Communication and Conversationality: This characteristic is one of allowing for a

multimedia multi-user environment that permits a full sharing of the environment a an equal
basis amongst all of the users. It further allows the users to interact with any other user while at
the same time allowing this interaction along any one of the multimedia dimensions.
3. Interactivity and Responsiveness: This dimension relates to the ability of the environment
to allow one or several users to utilize all elements of the medium and at the same time to pose
questions that are robust in a multimedia sense and to obtain adequate answers.
4. Presentation and Interaction: The interactiveness of the environment is a key element of

understanding the
5. Non-Linearity and Hyper-Dimensionality: This dimension of characterization allows for

the movement amongst the object in an unbounded fashion. It allows for movement in space, all
dimensions, and time, as well as in point of reference. The spatial movement entails the ability to
view at different magnifications that is common amongst hypermedia environments. The ability
to view at different points of reference allows one users to accept the reference frame of another
to view the object.
6. Sense-Complexity and Representation: This dimension allows for the combining of

multiple sensory elements into the multimedia objects as well as the presentation of those
elements either as direct manifestations or as appropriate analogs.
In the context of this paper there will be three elements that define multimedia and multimedia
communications; the message, the medium and the messenger. The ultimate result is the impact
of the information created on the environment. Without the result, there has been no transfer of
information. The objective of this paper is to show the relationships between all three of these
elements and the blend with them a set of philosophical underpinnings that will allow them to be
used in analyzing the development of multimedia communications.
48 | P a g e
We first define the three elements of message, messenger and medium.
1. Medium: The medium is the collections of all physical elements outside of the mind of
the creator of the message that facilitates the externalization of the message. Paper, a video
screen, a hypermedia environment, a set of signaling flags, a stone tablet are all the elements of
the medium.
2. Message: The message is the "idea" to be transferred from one individual to another. It is
the information content to be transferred and thus to acted upon. It is an actionable element of
internalized conceptualization.
3. Messenger: The messenger generally is thought of as an individual. In our context, the

messenger is the collection of any and all entities that move the message from one point to
another. Recall that the movement may be i space and time. Recall also that the channel used in
the movement may be a "noisy" channel that can introduce errors
There are three schools of philosophical thought that will be used to assist in developing and
overall theory of multimedia communications and these schools, complementary in many ways,
assist in each of these three elements. They are;
1. Semiotics: This is the study of "signs". In the current context, semiotics helps us with the
study of the medium elements of multimedia communications. For indeed, the medium consists
of the signs that are used in the conveyance of the idea. To the semiotician, the world is viewed
in the sense of a pure sign, an outward display that can be interpreted to mean what the displayer
intends. The semiotician reflects on the issues of identifying the sign and then identifying the
meaning.
2. Deconstructionism: This is the school of thought that tries to understand the idea or the
essential information from the message. What was really meant by the message sent. It takes the
combination of the message, the medium and the messenger, in context, and tries to determine
the essential message or information content. It does this through the inherent assumption that
messages are actionable and it is through the direct of implied actions that the information may
be revealed. The deconstructionist relies upon the deeper philosophical underpinnings of the
conveyer of the sign. To properly deconstruct a sign, the deconstructionist must understand what
world view or philosophical sets of precepts the holder of the sign has. Deconstruction begs the
question of a real philosophical ethos, expressed or not expressed.
3. Hermeneutics: This is the study of what the messenger really meant by the words uttered.
Hermeneutics focuses on the message via the messenger. It is an extension of biblical study of
49 | P a g e
trying to reinterpret the message of the gospels and old testament writers from the human
perspective of the writer, in our case the messenger.
Our approach is to define and the describe each of the elements in the above and then to apply
them to multimedia communications. The application will focus on two issues. First we will
develop a set of analytical tools that will allow the user to understand an existing multimedia
environment. Second, we will develop a methodology to synthesize the multimedia environment
that meets the objectives of the designer. The reader must note that it is the joining of the words,
the ideas and the signs that leads to the actionable event.
This paper is structured to address each of these above issues. It first focuses on the message, the
medium and the messenger. It then shows how the three philosophical schools ca converge to
better understand the issues of multimedia communications.
2.3 MULTIMEDIA; ELEMENTS AND STRUCTURE
Multimedia technology or systems have been viewed as multisensory end user focused
technological add-ons to existing or new computer processing environments. Multimedia is a
catch phrase for new technology that displays, either orally or visually, information that may be
stored in an dense and complex form at the end users site. It has been viewed in many ways as a
means to an end, yet viewed as an end in itself. To date, there has been no "philosophy" of
multimedia, no Marshall McLuhan of the multimedia world to articulate who wants to use it and
for what purpose. The key questions from a business perspective is how does the set of
multimedia technologies provide increased value to business customers, and how does the
consumer value the multimedia technology in their daily lives. The use of multimedia technology
and its acceptance is an economical factor, based on value creation and perception on the part of
the user. It is not, as has all too often been suggested, a technologically driven market.
2.3.1 Elements of Multimedia
To begin, the elements of multimedia as currently understood by the hardware and software
providers can be presented. There are six major categories of multimedia technology and
elements. These are;
1. Displays: The MM display areas focuses on enhancement of images, improved

resolution, and display presentation processor enhancement. The display may be generically for
any one of the senses; sight, sound, touch, or others that may be appropriate. HDTV is an
example of one area of progress. Especially digital TV.
50 | P a g e
2. Processors: The processors and processes are the special software or hardware elements
that locally enhance, manipulate, display, or interpret multimedia elements.
3. Content: This is the more general software that is created and used on a multimedia
system. It is the result of authors works in creating information and interactive systems for
multimedia.
4. Storage: This is the composite of elements that are used for a multimedia database,
storage and retrieval system.
5. Enhancement: These are special purpose non-co-located processors and processes that are
used for special purpose enhancements.
Communications: This is the ultimate heart of a multimedia system, namely the ability to
communicate in a full and complete multimedia fashion. This is a complex amalgam of
protocols, transport infrastructures, servers, and interfaces that allows and support multimedia
communications amongst a group of users. All other elements are at best users to interface or
utility, and not user to user. Multimedia communications systems support sessioning; namely
synchronization and orchestration of multi-media/multi-user conversationality.
These elements must be combined in a fully integrated and networked fashion to deliver a
transparent system or service to as wide a variety of end users as is possible.
2.3.2 Value Creation in Multimedia
The use of any new technology must fit within the overall context of value creation to the user.
Value, as a concept, may be related to the microeconomic concept of the utility function of
demand but it more closely is related to the ability to market the new product or service directly
to the customer. Namely, value in the context of multimedia communications is best defined in
being able to quantify for the user what savings in expenses or capital shall accrue from the use
of the new product or what new revenue stream will result. Ultimately, value in an commercial
context is nothing more than the increase in the net present value of the business entity. In the
consumer market value is measured by means of its ability to displace other expenditures that the
consumer already has made for a perceived greater value from the product offered.
In either case, value creation is essential when determining what multimedia has to offer. The
focus in this White Paper is primarily commercial and not consumer. The latter has a greater risk
threshold in uncertainty and will be dealt with latter. In the commercial context, therefore, value
creation is a definable and measurable result of using the new technology in an existing business
context. The business imperative, therefore, is that multimedia services must create value for the
firm.
51 | P a g e
2.3.3 The Multimedia Food Chain and Its Missing Links
The lack of understanding of what multimedia is, combined with the need to meet the measures
of the customers value chain, dictate that all elements of the customer's operational environment
must be understood and considered to effectively establish and operate an effective multimedia
service business. This is called the "food chain" concept. Having even a single element missing
will result in starvation no matter how robust the other elements in the chain are. Thus, it is
essential to determine if, in the development of a service business, any one company, supplier,
customer, or other such entity, has been left from the flow of the service and thus will cause it to
fail.
For example, in the healthcare market, imaging and multimedia applications have been
developed in significant numbers. However, several key elements have been missing.
Specifically; bill transaction processing and record management and keeping. The current
approaches to the delivery of medical imaging systems attempt to support applications in
radiology by merely replacing the viewing screen. However, this is but one step in the process.
AT&T and Philips have learned the hard way that such a point replacement is unacceptable. The
system that they have developed, COMVIEW, has had limited acceptance. The system is a
typical high end multimedia system. It handles complex images and text, allows for the
integration of voice into the overall system, and provides a limited amount of record
management. However, it does not readily fit the pattern of the radiological suite in most
hospitals. It does not solve the integration of record management and does not solve the issue of
patient record keeping and billing.
In the Hospital environment, the use of multimedia will be driven by the need to treat each
medical Department as a profit and loss center. Revenue must be ascribed to each procedure and
each patient and expense tracked. Quality care will be an equal imperative. Thus a multimedia
system must be one that starts with that premise, allows for graceful and incremental migration
and addresses the needs of the physician, the technician, the nurse, the administrator and the
support staff. The same can be said about all other market applications. Moreover, multimedia
systems and services in healthcare and other similar markets, will have a significant impact
because these markets are very information rich, and require communications of this information
to many people. It is this nature of information richness, in both type and form, combined with
the need to transact along with the sharing of the information that establishes a need for
multimedia services.
2.3.4 Elements and Structures
The end user interface in a multimedia environment is dramatically different than that in a
traditional data processing environment. Specifically, in a multimedia environment, we have
52 | P a g e
more intimately involved the human as a processor and evaluator of the information flow and we
furthermore have complex information sets flowing from one point to another as well as in a
conversational fashion. We will develop the end user concept in terms of the paradigm
developed by Winograd and Flores.
In the development of a theory for design of computer systems involving the human user,
Winograd and Flores invoke the theories of the German Philosopher, Heidegger. Specifically
they refer to four key propositions of the philosopher that impact the overall end user interface
issue in the multimedia environment. These are:
Our implicit beliefs and assumptions cannot be made explicit.
We all too often may make the statement, "You know what I mean." In so doing we are creating
to mistakes. First, the other may never know what we mean just by the nature in which we
individually perceive experiences and objects. Second, we may, ourselves, not have the insight to
our own true beliefs, because we all too often find ourselves questioning them. Hermeneutics,
the study of meaning in documents, has been expanded by Gadamer to investigate human
reasoning. Thus, indicates Gadamer, our understandings can change with the time and place.
This changing makes the explicit articulation specious at best.
Practical understanding is more fundamental than detached theoretical understanding.
Heidegger has a concept called "throwness", part of being-in-itself. We know something only by
being thrown or involved in it. We know what a radiologist does with an image and how he
manipulates it for understanding by doing the process ourselves. We cannot expect the user to
detail their beliefs and in fact those understandings are time varying.
We do not relate to things primarily through having representations of them.
We relate to things themselves. We do not relate to a representation. The representation to the

"thing itself" is done in the context of the task to be accomplished. For example, teleconferencing
is useful is we are not to relate to the person but to a subject whose essences can be presented
directly through the medium, rather than just a representation. We find that teleconferencing is
inadequate for personal contact since the contact is through a representation.
Meaning is fundamentally social and cannot be reduced to the meaning giving activities of
individual subjects.
53 | P a g e
Meaning is obtained in dialog, in a conversational fashion, with the ability to meet consensus.
Gadamer and Heidegger both relate meaning to the social process of communicating. Both also
relate the evolution of meaning to the ongoing set of discourses.
Specifically, social or conversational activity is the ultimate foundation of intelligibility. This

means that both in the design process as well as in the operations process, the need is critical to
have the communications channel be conversational if the intent is to convey intelligibility. If the
intent is only to transfer predefined package from one point to the other them the
conversationality is not essential. In a multimedia environment, intelligibility in the context of
the various media and thus intelligibility demands conversationality.
We can also try to better understand the interface by recognizing that the challenge is matching
man to the machine. To do this we are frequently pressed into a metaphorical set of analogies.
Typical is that of "man as the human computer". Metaphors of this type are both powerful
explanations of new concepts and clear statements of our total lack of understanding of the issue.
To quote from Mac Cormac:
"Explanations without metaphor would be difficult if not impossible, for in order to describe the
unknown, we must resort to concepts that we know and understand, and that is the essence of the
metaphor, a juxtaposition of the familiar and the unfamiliar."
MacCormac further quotes from Arbib, Man a Machine, The Metaphorical Brain;
"We want to understand how people think and behave....In some ways the brain of a man is like
the computer of a robot, in others it is akin to the brain of a frog. Our aim here is to convey an
understanding of the brain in terms of two main metaphors: The cybernetic metaphor, "Humans
are machines," and the evolutionary metaphor, "Humans are animals." We shall downgrade the
differences, but we hope to learn much from the similarities."
The harshness of Arbib pronouncements are striking. For indeed he represents the voice of many
computer designers who view the human at best as a fellow computer and at worst a level above
slime mold. Metaphor is powerful for it is in essence our most gracious way as humans to
express our total ignorance of the true essence of the problem at hand. To use the Heidegger
view, the use of metaphor is essential if we have not experienced the throwness of the problem,
that we have not immersed our total being in the basic uncertainty at hand. We use metaphors as
a way of re-expressing what we already know rather than understanding the unknown.
The danger in the use of such metaphors is clearly that we fail to come to deal with the needs of
the end user in interfacing with the multimedia world. We view the end user as another
peripheral computer system and not as an entity that must be thrown within the environment to
54 | P a g e
best profit from its performance. Thus, as we saw in the last section, current authors view
multimedia as nothing more than another display or another storage device. Their world view, as
so aptly developed in Kuhn's thesis, is limited to the existing paradigms. They talk
metaphorically as man as computer, or worse as man as frog.
This world view is dramatically different from that of the rational school of thought that focuses
on the idea that there exists a perfect truth independent of the individual and that through proper
perception as a single individual this truth can be made clear. Heidegger's approach is that we
must combine the rational objective world with the totally subjective individualistic world into
an environment where the human users becomes part of the environment of the media.
As Gadamer has stated (see Warnke), we understand in a dialog manner. Specifically:
"If one examines Gadamer's analysis ...all knowledge of the natural and social world...is
grounded in traditional orientations. We never come upon situations, issues or facts without
already placing them within some context...and interpreting them in some fashion."
"In equating the logic of understanding with the structure of dialogue, Gadamer suggests that the
proper answer is that..in genuine conversations ...all participants are led beyond their initial
positions towards a consensus.."
Thus the process of consensus in a conversational mode is what leads to new understanding. All
initial constructs are based upon prior prejudices that can best be formed in the context of
metaphors. If our goal in developing new user interfaces is the ability to allow the users to
understand, as viewed by Gadamer, then we must do so as to support the conversational modality
and to allow the reaching of consensus.
Winograd and Flores have noted six effects of accepting the Heidegger world view. These are;
You cannot avoid actions.
Even inaction is a form of action. Managers, as developed by Simon, interact with their day to
day industrial environment, and managers who act by inaction have the corresponding results.
You cannot step back and reflect.
Events exogenous to us are continually occurring and any attempt to stop time to best understand
the situation is at best specious. At worst, it becomes inaction. The concept of hermeneutics is
one that extended to the environment of the end user say that we make interpretation with what is
at hand and what is part of our tradition.
55 | P a g e
Effects of actions cannot be predicted.
We can anticipate, we can plan and we can strategize, but the world is filled with uncertainty. As
such, we act in an environment where the exact outcome is uncertain. The user must anticipate
that but not be fearful of it.
You do not have a stable representation of the situation.
Every situation is a representation in flux. When a user accesses a system, there are many factors
that impinge on the interaction of the user, their needs and responses. No system interface to a
user should assume a stable representation of facts. Designs should be such as to prepare for
ambiguity.
Every representation is an interpretation.
X rays are inherently representations of physiological factors. In looking at an x ray a physician

is looking at a representation and performing an interpretation. When we design a user interface,
we are representing a set of facts to the users. The act of the designer in representing the facts is
an act of the designer in interpreting for the user the facts. Thus in designing the interface, the
designer must be aware of the fact that they are entering into the interpretation process. Not only
is the user interpreting but so too is the designer for the user.
Language is action.
Speech through our language is a spontaneous reaction to a set of situations. In the design of
computer interfaces we spend many hours on structuring the presentation of the visual material.
Images are carefully scrutinized. Speech, in a multimedia context is fluid and open to instant
interpretation that may not be consistent with the other participants in the multimedia session.
For example, our tone of voice may make us appear arrogant, our questioning may make us
appear petulant and our suggestions may make us appear pedantic. Despite all our structured
work on the interface, the instantaneous impacts of the language may override the setting. Thus a
multimedia environment must have the flexibility to self-correct.
There are eleven design guidelines that Winograd and Flores have articulated and these play well
into the end user interface effort associated with multimedia system. These guidelines are as
follows;
56 | P a g e
1. There are no clear problems to be solved. Action needs to be taken in a situation of
irresolution.
2. A business is constituted as a network of recurrent conversations.
3. Conversations are linked in a regular pattern of triggering and breakdown
4. On creating tools were designing new conversations and connections.
5. Design includes the generation of new possibilities.
6. Domains are generated by the space of potential breakdown of action.
7. Breakdown is an interpretation - everything exists as interpretation within a background.
8. Domains of anticipation are incomplete.
9. Computers are tools for conducting the network of conversation.
10. Innovations have their own domains of breakdown.
11. Design is always already happening.
If we follow these design rules in developing the human interface and if we understand the
underlying theories of human understanding and intercommunication, this will assist the designer
in being flexible to converge to a more stabler interface.
Language is a means to expresses knowledge. Language is also a means to gain and crate
knowledge. We can now expand the concept of a language from what we see as words and what
we hear as speech to what we see as actions and what we create as situations. The ability to
provide for more breath of language as intercourse and interaction, and the ability to extend that
intercourse to all of the senses, no just sight and sound, allows for the attainment of the
fulfillment of a multimedia communications environment.
The concept of the end user interface that we developed in the last section centered around the
need for conversationality. Conversationality is embodied in the concept of a session which is
the electronic implementation of communications in a multimedia environment. We shall discuss
the concept of a session in more detail in Section 5 on communications. Simply put, however,
the session is the underlying communications construct that ensures the binding of users
57 | P a g e
together. Understanding the multi-user environment will provide the boundary conditions that
envelop the communications environment.
The essence of multimedia communications in a multimedia environment is the embodiment and

completion of transactions. Transactions are an ordered set of actions taken by the set of
participants in the session, whose completion leads to the successful
The users in a multimedia world are categorized into several classes. We can consider three
levels for discussion purposes and we may possibly expand these as we proceed. The three are
the end user, the physical and the virtual user. The end user is a definable entity that has action
responsibility for effecting the transactions that occur in the system.
Data elements in a commonly accepted database are collected according to some schema and
labeled accordingly. For example, we may consider a typical database containing the elements of
name, address and phone number. In a computer database these elements are then encoded into a
digital format and the name, address and phone are stored in some binary form for latter
retrieval. The retrieval process can be performed by asking a simple question of the database,
specifically, "List all of the name for the case where the zip code is 05XXX." Here we have
asked the database to perform certain acts. First to go through all of its records and perform a
match on the basis of zip codes. This is a redials performed task since all it entails is matching a
bit pattern for the desired code with the bit pattern for each data entry along the zip code field.
We know how to do this. The we accumulate all the names and finally print them out.
There are certain inherent structural assumptions that we make in this type of database. First we
assume that each data element of a data object has a defined structures. Second we assume that
each data element is decomposable in a finite sequence with uniquely defined data objects. For
example, we know that address is composed of the objects of street number, street name, city,
state, and zip code. We further know that zip code is composed of five digits, not characters. We
also know that there is no other representation for address and that further the object address
contains no other information. Address, for example, cannot tell me about the type of house, its
color or the number of windows. The questions that I ca ask are a priori stated and implicit in the
structure of the data object.
Now consider a multimedia data object. The object is an image, specifically an x-ray. Now we
can create a patient record which contains the information of the type, name, address, date of
admission, attending physician, blood tests, and x-ray. The objects, with exception of the last in
this record, are of the type that we discussed above. They are bounded, fixed, and definable.
Specifically we can ask a specific question and obtain a quantitative answer. In contrast we
cannot ask questions of the x-ray on a specific basis and hope to get an unambiguous answer. We
58 | P a g e
may ask for the last name and get the answer "Jones". We can ask about the x-ray the question of
the disease and get a totally ambiguous answer.
A more common data object may be considered to be a self-correct hyper object if we allow for
the depth that may be part of the decomposition of the object. For example, we have the record
for "Student" which is composed of name, address, and grades. Name is further decomposed into
last, firs, middle, and even parents. Parents is decomposed into mother and father.; And they in
turn into last, middle, and first. The common data object is in itself a hyper object having depth
and extent. However, the depth and extent has been defined a priori and is part of its very
creation. The database designer had defined fields for allow of the elements and for the
relationships between the hyper objects depth and co-relations.
If we extend the hyper object case to the multimedia object, we fond dramatic differences.
Consider again the case of the x-ray. At best the database designer assigns 100 million bits of
space for the data object. There is no structure. For example, the image may be that of a chest x
ray and we may be interested in the lungs, the heart, the stomach, or any one of several organs.
In the context of that interest, we are further interested in asking a set of important questions as
to the nature of the lung's clearness, and if not clear, what is the nature of the perceived
shadowing. All of these are useful for a careful diagnosis. Thus with a multimedia object, the
object contains information in a complex form, a form that is processed by the cerebral cortex
only a posteriori, and not a priori. Further, the information has dimensions that are not fully
known before the interaction with the end user. This is the essence of the concept of Heidegger
that we had discussed as an integral part of the Winograd and Flores theory of computation. A
multimedia object is truly a new paradigm for the representation of data.
The record in a standard database is quasi static in nature and can be changed in a typical
transaction processing system only upon the commands of the central data base administrator or
upon the allocation of commands in a distributed database environment. Thus, for example, we
can envision a database that has the record for the number of seats available on a typical airline
flight. The seats available may be changing with time but the change is controllable and is
synchronized by some overall database mechanism.
In a multimedia object, there is change in the object that is a natural progression of the object in
time. For example, if the multimedia object is a speech signal, we know that if the signal was
sampled at 64 Kbps then we must play the record back at the same speed. If we create a
compound multimedia record of voice and video, we must now synchronize not only within the
objects but also between the objects. Time now plays an integral role in even static records. If we
now extend this to the case of a dynamic transaction like multimedia record, we find that the
maintenance of the synchronization within and between objects is as critical an element as is all
others.
59 | P a g e
Thus multimedia objects have a complexity and richness of structure and dynamism and
interrelationship of form that separates them dramatically from the typical data records that we
use as part of the typical data environment.
Multimedia objects have three major dimensions of complexity. They have a temporal, spatial
and logical structure. The temporal structure there is tempo to the object as there may be in a
speech segment or a video segment. The tempo may also relate to the movement of a cursor and
its relationship to other objects. In addition, the temporal object structure may be a segment
(namely bounded a priori such as an image) or a stream (uncertain terminus such as a speech or
video segment). The temporal structure of a multimedia objects reflects the timing, sampling
rate, type of object, appropriate delivery time and the boundedness of the object.
The spatial characteristic reflects the ability to decompose the multimedia object into spatial
parts. This is particularly true of images that have some spatial dimension in two or three
dimensions. The basic need her is to have a decomposition structure that can be discussed about
the object in an in band or out of band fashion. The metaphor of boundedness is a concept that
reflects the location of the information on the spatial decomposition of an object. If the object
has its rules for decomposition imbedded in the data stream itself, we call it in band. If the
decomposition information is located as a separate descriptor, namely a data header, we call it
out of band. We must deal with this fact. In images, we restrict the boundedness to a finite set of
display elements. Thus we know a priori that an image is 2,000 by 2,000 pixels of 24 bits per
pixel, rather than an unknown length voice segment.
In the logical domain, we also are concerned about decomposing images. Logical decomposition
may be simple to state but very difficult to implement. For example we may want to decompose
objects in the form of type of bones in x-rays or by excited speakers in a voice segment. There is
currently limited analytical processing power to extract this type of decomposability.
We can then further combine these simple multimedia objects together into a compound
multimedia object. The compound object now has needs for orchestration, that is timing all of
the timings of the simple objects. The concept of orchestration is simply just that, being the
conductor to assure that all of the instruments in the symphony are not only properly tuned by
timed with regard to each other. We also worry about the issue of concatenation, the opposite of
decomposition. Concatenation states how we handle the spatial and logical combing of the
simple objects.
Communications in a multimedia environment is a process of allowing agents of the

communications process at various levels to interexchange information in support of the
complete end to end flow of multimedia elements. This definition allows for the existence of
60 | P a g e
agents which are definable and locatable entities that have specific responsibility for
communications flow and management. In the world of multimedia communications, these
agents must be empowered with significantly more responsibility than is typically the case in a
data communications world.
We will develop the concepts of the communications environment in the multimedia world by
first developing the concept of a session. A session is a shared multimedia conversational mode
of communications allowing multiple users and devices to share a common working space in
such a way as to make the communications interfaces and transport not only transparent but
acting as a facilitator of the communications process.
A session provides for several key layers of functionality. These are:
Event Management: Any multimedia activity leads to a transaction. The event manager is the
transaction manager. Multimedia events are complex and thus the manager must deal with these
distributed complexities.
Dialogue Management: The multimedia conversation must allow sharing of resources and must
encourage and facilitate conversationality. Dialogue management is that function.
Activity management: Activities are extended events, they are displaced conversational elements
that must be remembered and connected.
Synchronization: The orchestration of the many elements in a multimedia conversation is the

role of the synchronization function.
We must define entities that connect together and further define the entities that ensure that this
session level connectivity is supported.
2.5 Definition of Multimedia Services
Multimedia Services requires a working definition. This section attempts to provide such a
definition. It differs dramatically from many of the definitions offered for multimedia in many of
the current trades or even in the business and strategic plans for companies purportedly in this
business. The following is the definition:
Multimedia Services is a set of services offered to a community of users that enables and
empowers them to perform tasks in a collective fashion that can be accomplished in a
significantly more productive fashion by including multisensory information and interexchange
in a fully conversational mode, transparent to any and all users, and allowing all of the processes
61 | P a g e
performed to be monitored, recorded, retrieved, and transacted in a fully electronic fashion.
Multimedia Services responds to the needs of the users and can be measured in the context of
increasing value to the user base and can be provided in such a fashion that it does not neglect
any element of the organizational food chain.
Multimedia Services has certain characteristics that empower it to enable the user to achieve the
value that has been determined. Multimedia Services must account for the elements of the human
and organizational environment. From the human perspective, several of the common features
are as follows:
2.5.1 Human Factors
1. Conversationality: A system, as service, must allow, and more importantly encourage and
support, a conversationality amongst users that enables all of the multimedia senses.
Understanding and meaning are essential a social act that requires the full sensory interaction
amongst individuals and permits them to act together. Multimedia services must not only enable
this type of activity but must do so in a fashion that is consistent with existing social paradigms.
The essence of multimedia conversationality is its ability to empower the user to be "present-at-
a-distance".
2. Throwness: The ability of the user's actions to have some immediate or consequential
effect on their environment has been called being thrown into the environment and the term
throwness has been used. This capability implies that the user of the service is able to manipulate
the object or objects at hand in all of their sensory dimensions. The use of a multimedia service
in the design process must be such as to allow the designer to see the impact of the design in
terms of its tactile and visual elements, and also the aural elements that may be part of the
design. The designer must also have the ability to "use" the designed object in the context of its
use to better understand it functionality and acceptability.
3. Breaking Down: The individual does not understand something of an object, especially a
multimedia object, unless they have the ability to break it down, namely do something with it to
make it function as we know it. Take for example the use of multimedia in healthcare pathology.
When a physician views a slide of a tissue sample, they are viewing it after having done so many
prior times and having related certain cell boundaries and shapes, blurs to the uninitiated, yet
critical to diagnosis of certain conditions of the body organs.
4. Transparency: A multimedia service qua service must be transparent to the users. For
example, setting up a simple conference call on a telephone is made so by calling an operator
who then calls all of the participants and adds them to the conference. Cumbersome as this may
62 | P a g e
be it creates a sense of transparency, namely, no one user must know how the call is to be set up.
Multimedia services must be operator independent and yet have transparency to operations. This
will be one of the major challenges of the service. It will also be a barrier to exit in a competitive
market.
5. Action: The service must allow and enable action. This may mean that in a multimedia
session, control of the conversation can be taken around the session, allowing any single
individual to explain a point through demonstration. It may mean the obtaining of an object or
element on demand, and then having the ability to manipulate it at will. It is the ability to
demonstrate, contradict, confer, and compromise by example.
The second set of factors to be part of a multimedia service relate to the organizational structure
that the service fits into. Specifically:
2.5.2 Organizational Factors
1. Transaction: The service must support the total set of transactions that will occur through
its use. At the simplest end, it must create and process bills for services rendered as part of the
service. It also must track each interaction of a user with the system in order to support customer
service and assist customers in the event of problems.
2. Productivity: The service must address the needs of the customer for productivity
improvements. This means that the service must immediately reduce costs through more
effective use of existing manpower. For example, in the advertising application, the service must
reduce the cost and time associated with the development and production of ad copy. In short, it
must be cheaper and faster.
3. Infrastructure: The service must become transparent to the user. It must exhibit all of the
qualities of an infrastructure.
4. Value: The service must have a calculable value to the user. The service must address
specific processes or operations, with known cost structures, and must definitively show how the
costs are reduced and the value of the business unit increased.
5. Holistic: This implies that the service must integrate into the overall way in which the
business operates. It should not optimize one part and fail to address inefficiencies in others. It
must address the entity as a whole.
2.4 THE MESSENGER; COMMUNICATIONS AND INFRASTRUCTURE
63 | P a g e
The messenger is the process and processes that link with the medium to effect the transaction of
the message. It is of interest to see that the old AT&T used Mercury as the messenger. Hermes,
also a messenger, although a mischievous one, is a better choice. Mercury delivered the
messages correctly. Hermes would always put a little twist on them. We shall in this section
develop the concept of a communications system and then move forward on how that system
may be viewed as an infrastructure. A great deal of talk about infrastructure has occurred as of
late with little definition. We try herein to define out terms. This will be critical as we apply
philosophical principles to the issue of messages, messengers and media.
2.4.1 Communications Systems
There are four architectural elements in the telecommunications network. These elements are the
control functions, the transport function, the interconnect function, and the interface function.
We now provide further detail on these functions. It should be noted that these functions have
evolved over the years in content and complexity. We view these elements in the context of a
communications network that must support the most advanced current concepts in
communications. Specifically, the world view adopted in this paper that lead to an interpretation
of this architecture are:
(i) End users desire to have interactions in a real time fashion with images and other high
resolution information that must be provided in a fashion that meet both time and resolution
requirements (See Barlow).
(ii) The end user devices are extremely intelligent and complex and can operate in a stand-alone
environment.
(iii) The users desire to operate in a totally distributed fashion. Data bases will be a different
location, users are at different locations and input output devices are also at different locations
(See Dertouzos and Moses, and de Sola Pool pp. 57-59 for details on these directions).
(iv) The network may provide different levels of service to different users. There is no need to
provide universal service of full capability to all end users.
This view of the network will significantly influence how extensively we defined the elements
and in turn will impact the combination of those elements in an overall architecture. All of these
assumptions on the world view are different than before, in an all voice world. In this paper, we
define a network as an embodiment of an architecture, in all of its elements.
The architectural elements are control, transport, interconnect and interface.
64 | P a g e
1. Control: Control elements in an architecture provide for such functions as management,
error detection, restoral, billing, inventory management, and diagnostics. Currently, the voice
network provides these functions on a centralized basis, although in the last five years there have
evolved network management and control schemas and products that allow for the custom
control and management of their own network. Companies such as IBM, AT&T and NYNEX
have developed network management systems that move the control from the network to the
customer (See McGarty and Ball, 87, for a detailed discussion of the different types of control
and network management strategies). On the sub-network side, companies such as NET,
Timeplex, Novell, 3-COM and others have done similar implementations for local area
networks, data multiplexers and other elements. Centralized network control is now longer
necessary and in fact it may not be the most efficient way to control the network. What is
important, however, is that network control providing the above functions is an essential element
for either a public or private network. Thus as we consider network evolution, this element or set
of function must be included. Control has now been made to be flexible and movable. The
control function is probably the most critical in the changes that have been viewed in the context
of an architecture. All buildings need windows, for example, but where one places the windows
and what one makes them of can yield a mud adobe or the cathedral at Chartres. The same is true
of the control element. In existing networks, the control is centralized, but in newer networks, the
control is distributed and empowered to the end users. The users can now reconfigure, add,
move, and change their network configuration and capacity
2. Transport: The transport element is provided by the underling transport fabric, whether
that be twisted pair of copper, fiber optic cable, radio or other means. Transport should not be
mixed or confused with other elements of the network. Transport is merely the provision of
physical means to move information, in some form such as digital, from one point to another. At
most it is expressed in bits per second and at best it is expressed in bandwidth only. Bandwidth
as a transport construct is the most enabling. Transport does not encompass the need to change
the information or to do any other enhancement to the information.
3. Interconnect: The interconnect element of the architecture describes how the different
users are connected to one another or to any of the resources connected to the network and is
synonymous with switching. Interconnection assumes that there is an addressing scheme, a
management scheme for the addresses, and a scheme to allow one user to address, locate and
connect to any other user. Interconnection has in the past been provided by the Central Office
switches. As we shall discuss latter, this implementation of an architectural element was based
on certain limitations of the transport element. With the change in the transport element of
structures allowing greater bandwidth, the switching needs have changed. Specifically,
distributed systems and scale economies of the distributed Architectures allow for
interconnectivity controlled by the CPE and not the Central Office. As we shall show later, the
advent of Local Area Networks and CATV voice communications are ones using distributed
65 | P a g e
interconnectivity elements. There are three general views of interconnection that are valid today;
the Telcom, the Computer Scientist, and the User. The Telcom view is based on the assumption
of voice based transport with universal service and the assumption of the inseparability of
interconnect and control. The Computer Scientist view is based upon the assumption that the
network, as transport, is totally unreliable, and that computer hardware and software must be
used in extremis to handle each data packet. Furthermore the Computer Scientist's view of the
network is one where timeliness is secondary to control. The Computer Scientists view has been
epitomized in the quote, "Every Packet is an Adventure". This is said with glee, in that each data
packet is set out across the network and it is through the best of hacking that the Computer
Scientist saves the packet from the perils of Scylla and Charybdis. The third view is that of the
user, who is interested in developing an interconnect capability that meets the needs and
minimizes cost. This is minimization of both obsolescence and cost strategy.Thus an investment
must try to follow the curve. In a hierarchical view of interconnect, such as a large centrally
switched network, the changes occur once every few years. Thus the lost cost or performance
efficiency can become significant. In contrast, in an end user controlled environment, with a
fully distributed architecture, the lost efficiency is minimized as technology advances.
4. Interface: The interfaces are the end users connection to the transport element. The
interface element provides for the conversion from the end user information stream and the
information streams that are used in the transport form of the network. For example, the
telephone interface for voice is the analog conversion device. Moreover, we must also view
interfaces as being composed of not only the hardware that provides for the physical
interconnection with the end user but the software that assists in the logical interconnection. The
telephone hand set is merely the first physical step in interconnection. A second step included the
PBX which included software that allowed for additional features. However, these features are
frequently not used because of the user lack of acceptance.
We have divided the network elements into these four categories to demonstrate that there are
clearly four distinct and separable areas for growth and policy formation. Issues of regulation,
due to potential monopolist control are always a concern, but it will be demonstrated that in all
four there are economies in market disaggregation.
Understanding that there are several varying architectural designs allows one to better understand
that each reflects not only connectivity but also the world view.
2.4.2 Infrastructures
Let us extend the concept of infrastructure. We find that there is a great deal of discussion about
infrastructures, in both academic and governmental circles. Unfortunately, none of the
participants deem it appropriate to define what they are talking about, namely what do they mean
66 | P a g e
by infrastructure. We feel that such a definition is critical to developing multimedia
communications concepts, since the concept of an infrastructure will be at the heart of the change
process. In our context, an infrastructure is a shareable, common, enabling, enduring, resource,
that has scale in its design, and is sustainable by an existing market, and is the physical
embodiment of and underlying architecture. Specifically;
1. Shareable: The resource must be able to be used by any set of users in any context
consistent with its overall goals.
2. Common: The resource must present a common and consistent interface to all users,
accessible by a standard set of means. Thus common may be synonymous with the term
standard.
3. Enabling: The resource must provide the basis for any user or sets of users to create,
develop and implement any and all applications, utilities or services consistent with the
underlying set of goals.
4. Enduring: This factor means that for an infrastructure to be such, it must have the
capabilities of lasting for an extensive period of time. It must have the capability of changing
incrementally and in an economically feasible fashion to meet the slight changes in the
environment, but must meet the consistency of the world view. In addition is must change in a
fashion that is transparent to the users.
5. Scale: The resource can add any number of users or uses and can by its very nature
expand in a structured form to ensures consistent levels of service.
6. Economically Sustainable: The resource must have economic viability. It must meet the
needs of the customers and the providers of the information product. It must provide for all of the
elements of a distribution channel, bringing the product from the point of creation to the point of
consumption. It must have all of the economic elements of a food chain.
7. Physical Embodiment of an Architecture: The infrastructure is the physical expression of

an underlying architecture. It expresses a world view. This world view must be balanced with all
of the other elements of the infrastructure.
An infrastructure is built around the underlying architecture. An infrastructure is in essence the

statement of the architecture which in turn is the conceptual embodiment of the world view.
Infrastructures as physical embodiments of architectures must, to have economic lives that are
meaningful, be developed when the world view, technology and user needs are stable. If any of
67 | P a g e
these three are in states of significant flux, the infrastructure may soon not meet the change in the
world view and then become obsolete.
It is important to distinguish between architecture and infrastructure. We have extensively

defined architecture in terms of its three parts: elements, world view and technology.
Infrastructure unfortunately has been reified in terms of some physical embodiment. The
discussion of NREN being an infrastructure is viewed by many as being a determinate thing.
Kahin has, however, de-reified the concept in terms of its being an embodiment of a concept or
set of common goals. We expand that and state that an infrastructure is an enabling capability
built around a common construct.
There are four types of infrastructure views that are pertinent to the current discussions of
networks. These are of particular import to such networks as NREN since they will lead to the
policy directions that it will take. These four infrastructure types are as follows:
1. Physical: This is the most simplistic view of an infrastructure. It requires a single

investment in a single physical embodiment. The old Bell System was such an infrastructure.
The National Highway system is such an infrastructure.
2. Logical: This network may have separate physical embodiments, but all users share a
common set of standards, protocols and other shared commonalties. All users have access
through an accepted standard interface and common higher level transport facility. IBM had
attempted in their development of SNA in the mid 1970's to develop a logical infrastructure in
data communications. This was expanded upon by the ISO OSI seven layer architecture,
selecting a specific set of protocols in each layer.
3. Virtual: This type of infrastructure is built on intermediaries and agreements. It provides

shared common access and support interfaces that allow underlying physical networks to
interconnect to one another. Separately, the individual networks may use differing protocols and
there are no common standards. The standards are at best reflected in the gateways to the
interconnection of the network. Thus this infrastructure is a loose binding through gateways. It is
in many ways what is the INTERNET today, if we include all of the subnets.
4. Relational: This type is built on relationships between the network parties and the
establishment on higher level accessing and admission. Specifically, a relation infrastructure is
based on agreements on sharing addresses, not necessarily common addressing, and on the
willingness to share data formats and types. It is an infrastructure based on shared common
interests but not shared common access. This type of infrastructure is what in essence exists in
most cases today. Users can move from network to network through various gateways. The
difficulty is the fact that the interfaces are cumbersome and may require sophistication on the
68 | P a g e
part of the users. However, more intelligent end user terminals and interfaces will reduce this
cumbersome interface problem.
Our conclusion is that understanding the type of infrastructure that the coalition of users want,
will also impact the architecture, based upon an imputed world view. Arguably, a physical
infrastructure leads to maximum hierarchical control and the resulting impacts that such control
leads to. This is a critical issue for networks such as NREN, since by choosing infrastructure and
architecture may not be as uncoupled as desired. In particular, the selection of Gbps capability
may really be GHz capability and is best suited to a Virtual or Relational infrastructure.
2.4.3 Current Infrastructure Options
There is a considerable amount of effort to define and implement an information infrastructure.

In this section we describe some of these current proposals, many of which are still quite
formative and lack substance. In some case we shall try to place them in the context of the
constructs that we have developed in the preceding sections. In order to fully describe these
infrastructures, it is also necessary to deconstruct the work of the authors, understanding their
meanings in the contexts of what they are saying and taking an approach that blends the
hermeneutics of Gadamer with the semiotics of Levi-Strauss. We now address several of the
more current views of infrastructures. In each case, we describe as best can be done the concepts
of each of the individuals, and then attempt a deconstruction in terms of their underlying
architectural assumptions, their view of infrastructure and more importantly their world view of
information and information networks.
(i) Dertouzos Infrastructure
This is the most widely discussed of the information infrastructures having been proposed by
Professor Dertouzos who is a Computer Scientist and the Head of the Laboratory for Computer
Science at MIT. Simply put, he defines the information infrastructure as:
" Common resource of computer-communications services, as easy to use and as important as

the telephone.."
Dertouzos states that there are three elements to his vision of an information infrastructure.
These are:
Flexible Transport: This includes bandwidth on demand, flexible pricing and security and
reliability.
69 | P a g e
Common Conventions: This includes his concepts of E Forms and Knowbots. The former is a set
of standard for formats and the latter are intelligent agents for the movement and processing of
data.
Common Servers: This is a set of common file servers or generalized servers to provide
directories, text/image translation, data base access and active knowledge.
In the paper, Dertouzos discusses this architecture and he uses as an example a system conceived
of and designed by the senior author (McGarty 1990 [1],[2], 1991 [1],[3]). In the author's system,
the assumption was to both empower the end user and to do so in an incrementalist fashion. The
architecture shown in this second system was based upon:
1. Available Transport: Take what is present and build in an economically viable fashion.
Build communications on an incremental and economically effective basis.
2. Open Interfaces: Use standards as appropriate, and allow the users the freedom to meet
their economic needs. Recognize the changing needs of the user and buyer and incrementally
change to meet the evolving needs.
3. Client Server Architecture: Maximize use of end user terminals and empower end user
applications development. Provide tools and not strictures.
The system designed and operated by the author actually connects the MIT campus with
hospitals, publishers, and other economic entities in a build-a-little, test-a-little, use-a-little
approach that allows for use acceptance and economic justification. The Dertouzos Infrastructure
assumes directions that are significantly different and diverge from the end user driven approach
of the author but take a more centralist approach. This latter approach has been advocated by
Moses in his discussions on the subject, yet are somewhat counter to Moses' layered
organizations that maximize flexibility and minimize complexity.
(ii) Kahn Infrastructure:
The vision of Bob Kahn, of CNRI is one of a broad band research backbone, loosely coupled,
with dark fiber and as high a bandwidth as possible, read data rate. This proposal, frequently
confused with the Gore infrastructure, is generally more open and flexible. However, it too lacks
any economic underpinnings.
(iii) Gore Infrastructure:
70 | P a g e
Gore, the Vice President of the United States in the Clinton Administration, son of the initiator of
the Federal Highway system, has argued for a single network, government directed and funded,
hierarchical in fashion, that allows everyone to have access to every bit. Consider his comparison
of data bits to corn kernels;
" Our current national information policy resembles the worst aspects of our old agricultural
policy, which left grain rotting in storage silos while people were starving. We have warehouses
of unused information "rotting" while critical questions are left unanswered and critical problems
are left unresolved."
He believes that every bit is a good bit. He further has no value concept of information. His
definition is clearly the one of quantity and not value. Researchers are not necessarily starving
for lack of bits. Quite the contrary, there is a need for coherent data reduction. He further states;
"Without further funding for this national network, we would end up with a Balkanized system,
consisting of dozens of incompatible parts. The strength of the national network is that it will not
be controlled or run by a single entity. Hundreds of different players will be able to connect their
networks to this one"
He is somewhat contradictory. On one hand he states that there should be one network and not
many, on the other hand he has all the separate networks connecting to this one. In this case, his
world view comes through clearly. He wants a hierarchical or at most centralized architecture as
well as a physical architecture. The proposal lacks the flexibility of an economic entity.
(iv) Heilmeier Infrastructure:
Heilmeier, the new President of Bellcore, the R&D arm for the Bell Operating companies on the
regulated side, advocated a hierarchical, BOC controlled, network intensive, monolithic network.
This is not surprising considering his extensive stay in Washington as a government bureaucrat.
He further argues for control of both wire based and wireless networks. He is quoted as saying;
"I'd like to see a bona fide information infrastructure rather than a fragmented world of different
systems for everything."
Networks are currently fragmented and as a result of this fragmentation local economic
optimization has occurred. In contrast to the hierarchical, centrally controlled view of Heilmeier,
also formerly head of DARPA, wherein he views the need for a single point of control and
direction, the world of communications networks and information networks have grown through
the increased power of the end user interfaces and interconnected distributed throughout the
network. In addition growth has resulted from less control in the network and less centralization.
71 | P a g e
The work of the author (McGarty 1990 [1], [2], 1991 [1],[3]) has shown an architecture for a
distributed multimedia environment that has been built and is still in operation that uses a mix of
communications channels and thrives on those channels that have the least functionality.
Specifically, dark fiber transport is the most enabling and empowering of any communications
channel.
(v) Proposed Infrastructure
Infrastructures are enabling entities. As we have discussed, an infrastructure does not have to be
a single centrally controlled, managed, and funded entity to be effective. In fact an infrastructure
on the loosely constructed basis of a relational infrastructure is just as effective as the extreme of
a physical infrastructure. We make the following observations, and based on the prior
developments in the paper propose an alternative direction for infrastructure development.
(i) Technology is rapidly changing and will continue to do so. The directions in technology
are towards increased processing capability per unit workstation and increased capacity in
performing both complex processing tasks while at the same time handling sophisticated
protocol procedures.
(ii) User terminals are expanding in a network multimedia environment that is empowering
the end users to both use many new media types as well as dialog in a conversational basis with
other users in the same network.
(iii) End users are becoming more pervasive and training of users based upon strict confines
of computer languages are disappearing. The end user is empowered to act and to use
information system with no training or education. Citibank, in its development of the ATM
network has ensured that the systems have minimal need for human intervention or training. In
addition, the Citibank home banking product, the most widely used of any home banking
products on PCs, is almost instruction free. The Apple MAC computer is also another example
of enduser empowerment through intimidation free end user interfaces.
(iv) Successful technology development in a productive fashion has best been effected within
the constructs of entrepreneurial small companies that allow for the creation of new ideas judged
by the dynamics of a free market. Large centralized technology development organizations have
time scales that are much longer than the time scales of the underlying technologies. The
developments in the computer industry of today are prime examples.
(v) Users are not only empowered to use systems in a variety of ways but they are also able
to select from a wide variety of systems, interfaces and data sources. To quote A.G. Fraser of
72 | P a g e
Bell Laboratories: "Every standards body seems to be churning out protocols left, right and
center. We may already have passed the point where we can all come together." (Coy, 1991)
Thus, distributed networks, interfacing with disparate other networks, through gateways is
already a reality.
These observations then indicate that with a changing base of customers, a changing set of needs
and an already progressing infrastructure that is relational at best, that to continue to maximize
our technical creativity it is best to match the information infrastructure to our cultural
paradigms. Thus it is argued that the proper evolution of an information infrastructure should be
along the relational model. That, in fact, the physical extreme is counter to the trends of user
empowerment and economic efficiency. It further could provide a roadblock to technical
creativity.
2.5 DECONSTRUCTION
Deconstruction consists of the unraveling of actions and physical realities to determine the
underlying sets of truth, such being definable and determinable. In its simplest state, the
unraveling is the determining of intentions within the context of how the intended views the
world at the time of the intent. Deconstruction from the perspective of multimedia
communications is a toll that helps the designer understand the biases, prejudices and limitations
of the designer. Deconstruction allows for the ultimate use of the technology in the context of the
user. We start the development of the deconstructionist approach with the development of the
concepts of paradigms and world views. We have already discussed them in some details but it
will be critical to the deconstructionistic approach to have them developed in full detail. Then we
shall move to the analysis of a set of key questions that appear when we begin the development
of a multimedia communications theory.
2.6 PARADIGMS AND WORLD VIEWS
The concept of an information system or communications architecture has been a cornerstone in

the development of new information and communications systems. However, the structural
elements of these architectures have not played a role in the development of policies. In this
section we will develop the concept of an architecture as a means to understand the network as
both a market and regulatory entity, and will provide a new set of perspectives for viewing the
network in terms of a new paradigms and world views.
The concept of a telecommunications architecture has been a cornerstone in the development of

new telecommunications systems. However, the structural elements of these architectures have
not played a role in the development of policies. In this section we will develop the concept of an
architecture as a means to understand the network as both a market and regulatory entity, and
73 | P a g e
will provide a new set of perspectives for viewing the network in terms of new paradigms and
world views.
An architecture, first, requires that the underlying system be treated in terms of a set of
commonly understood elements and that these elements have a clearly demarcated set of
functions and interfaces that allow for the combining of the basic set of elements. The way the
elements then can be combined, reflected against the ultimate types of services provided,
determine the architecture.
An architecture, secondly, is driven by two factors; technology and world view. Technology
places bounds on what is achievable, however those bounds are typically well beyond the limits
that are self-imposed by the designer or architect in their view of the user in their world. This
concept of architecture and the use of design elements is critical in understanding the paradigms
used in the structure of information systems (See Winograd and Flores, pp. 34-50, especially
their discussion of Heidegger and Throwness in terms of design). World view is the more
powerful driver in architecture (See Kuhn, pp. 72-85). We argue in this paper that it is essential
to develop a philosophical perspective and understanding of how to view networks. We argue
with Winograd and Flores, and in turn with Heidegger, that we must be thrown into the network,
to understand the needs of the users, and to understand the structure of the paradigms that are
used to construct the world view.
The concept of a paradigm is in essence the collection of current technologies that we have at
hand for the network and the ways we put these elements together. However, the true meaning of
a paradigm is in the context of the examples or experiments that we all relate to with that
technology. Paradigms are not technology in and of itself, but technology as example. New
paradigms result from new technologies. New technologies allow for the placing of the elements
together in new ways. Kuhn, then goes on to demonstrate that the world view, that is how we
view ourselves and our environment is based upon the our acceptance of these paradigms, as
either collections of techniques and technologies or as collections of embodiments of these
techniques and technologies in "examples". We then tend to accept this as the way things are and
should be. Then Kuhn argues, as the technologies change, changes in the paradigms do not occur
in a continuous fashion but almost in quantum leaps. The new paradigms build and congeal until
they burst forth with new world views. It is this model that we ague applies to the evolution of
broadband.
Thus, architecture is the combination of three parts: the common elements, the underlying
technology and the world view. We depict the conceptualization of architecture as the amalgam
of these three elements. We shall develop this construct more fully as we proceed.
74 | P a g e
The concept of a world view is an overlying concept that goes to the heart of the arguments made
in this paper. To better understand what it implies, we further examine several common views
and analyze the implications of each. If we view our world as hierarchical, then the network may
very well reflect that view. If we further add to that view a bias towards voice communications,
these two elements will be reflected in all that we do. The very observations that we make about
our environment and the needs of the users will be reflected against that view. As an external
observer, we at best can deconstruct the view and using the abilities of the hermeneutic observer,
determine the intent of the builder of the networks.
To better understand the importance of an architecture we develop the concept of the historicity
of architectures based upon the work of Kuhn and ten that of McLuhan. Kuhn begins his thesis
of how scientific revolutions occur by the introduction of the concept of paradigms. He defines
these as (see Kuhn p. 175);
"...the term paradigm is used in two different senses. On the one hand, it stands for the entire
constellation of beliefs, values, techniques, and so on shared by the members of a given
community. On the other, it denotes one sort of element in that constellation, the concrete puzzle-
solutions which, employed as models or examples, can replace explicit rules as a basis for the
remaining puzzles of normal science, the first sense of the term, call it sociological."
hand for the network and the ways we put these elements together. New paradigms result from
new technologies. New technologies allow for the placing of the elements together in new ways.
Kuhn, then goes on to demonstrate that the world view, that is how we view ourselves and our
environment is based upon the our acceptance of these paradigms, as either collections of
techniques and technologies or as collections of embodiments of these techniques and
technologies in "examples". We then end to accept this as the way things are and should be. Then
Kuhn argues, as the technologies change, changes in the paradigms do not occur in a continuous
fashion but almost in quantum leaps. The new paradigms build and congeal until they burst forth
with new world views. It is this model that we ague applies to the evolution of broadband. It is
this philosophical view, almost Hegelian in form, that is essential in understanding the
underlying and formative changes in paradigms that will change our world view.
As a second perspective of the impact of technology as a dominant driver, we can refer to

McLuhan and his development of the concept of media. Drucker has referred to the presentation
of McLuhan's doctoral thesis and McLuhan is quoted as follows (See Drucker, p. 250):
"Movable type, rather than Petrarch, Copernicus, or Columbus was the creator of the modern
world view.. "Did I hear you right," asked one of the professors as McLuhan had finished
reading, "that you think printing influenced the courses the universities taught and the role of the
75 | P a g e
university, altogether?" "No, sir, " said McLuhan, "it did not influence; printing determined both,
indeed, printing determined henceforth what was going to be considered knowledge."
This concept later evolved into the medium being the message. In our context it is the fact that
both Kuhn and McLuhan recognized, albeit in differing fields and in differing ways, that
fundamental changes in technology and technique, call it paradigm or the medium, will change
the world view, also the message. It is the importance of understanding the change in the
technology, its function and evaluate the possible change that this will have in the world view. It
will be argued, that much of the thinking in the current broadband areas, NREN in particular, is
based upon outmoded techniques and structures, and that a differing world view will evolve.
Take, for example, the use of twisted pair, pairs of copper wire, to transport telephone traffic. For
years it was implicitly assumed that this transport medium was limited to 4,000 Hz of bandwidth,
that necessary for an adequate quality voice signal. Specifically the world view was that of a
voice network that was to be used for voice traffic only. Ten years ago, this was a true limitation,
since the transmission was forcefully limited to 4,000 Hz by inductive loads or coils on the
telephone lines, assuring that you could do no more than the 4,000 Hz of bandwidth. Then, there
was a short period in the mid-1980s, when Local Area network manufacturers found that you
could transmit 1.544 Mbps over the common twisted pair, and that data was viable in what was
assumed to be a voice only medium. What had been almost religiously believed to be a limit was
found to be untrue. Then with the introduction of digital switches, the old "inductive loads" were
returned with the switch now limiting the data to 4 KHz or 64 K samples per second. The world
view of a voice only network took hold again, but this time in the context of a data rate
limitation, rather than a bandwidth limitation. In the early 90's there is another attempted break
out of the world view and to put 100 Mbps on twisted pair, so called FDDI circuits. Again, due
to the limitations on the part of the network as a voice dominated system, the world view keeps
this high data rate capability on the customer's premise only, and not the network.
We describe this transport world view evolution. Here we indicate the two dimensions of
information transport, bandwidth and data rate. The designer of the transport facility may limit
the data rate by selection of signaling format or delimit bandwidth by filtering. Twisted pair
actually has a bandwidth-data rate profile. It encompasses a large capability of either providing
bandwidth or data rates to the user. The two limiting world views are indicated as two solid lines,
one at 4,000 Hz and one at 64 Kbps. Both are voice only world views. We can readily see, that
with optical fiber superimposed the same issue of architecture dominated by world view may
result. In the fiber case, the result may be a segmenting of the architecture along selected data
rate lines, again formed by the voice world view.
76 | P a g e
Thus, architecture can be defined as the conceptual embodiment of a world view, using the
commonly understood set of nonstructural elements, based upon the available set of
technologies.
For example, Gothic architecture was a reflection of the ultimate salvation in God in the afterlife,
in a building having a roof, walls, floors, and windows, and made of stone and glass. Romantic
architecture was, in contrast, a celebration of man, using the same elements, but some employing
a few more building materials. The impact of the differences in world view are self-evident in
the embodiments of the architecture. (See the discussions on the impact of world view on
architecture in Wolfe. In addition see the cultural or world view impact on the Gothic
architectures in Jantzen and in Toy.)
Let us consider a second example of the impact of world view on architecture, specifically the
difference between the ISDN architecture and the architecture embodied in Local Area
Networks, LANs. ISDN is an architecture consistent with a voice dominated, hierarchical world
view of single points of control. LANs are architectures of world views that reflect both end user
self-empowerment and the environment of a data driven utility. This evolution in thought is
critical to understand the impact of world view. The LAN is an embodiment of empowerment of
the individual view, developed in the context of the 1960's and 1970's. The LAN concept,
originating at such locations as XEROX PARC, was driven by the developers needs to enable
and empower the end user with computing capabilities heretofore unavailable.
Out of this view came the LAN architecture of a fully distributed system, using a coaxial
transport mechanism to do nothing more than provide bandwidth. The transport mechanism is a
broad enabler. The actual implementation of the details is done at the users terminal in hardware
and software. This is in sharp contrast to ISDN, where the ISDN central switch does the
enabling. In ISDN, bandwidth is not provided, rather it is a voice based data rate, 64 Kbps or
multiple thereof. Consider this contrast in terms of how cable TV companies provided voice
communications in the early 1980's. Both Cox and Warner, using variations on LAN technology,
delivered a voice, video, and data service over the coaxial transport medium, by empowering the
end users terminal, not by regimenting the transport network.
Technology also plays a very pivotal role in telecommunications. Alfred Kahn (1971, p 300),
indicates that in the pre-divestiture period of the Bell System, the arguments for the needs of
both vertical integration and need for monopoly control were based on technology. Specifically,
there was a contention made by the Bell System that a single point of control to the network was
essential. Also, it was argued that an adequate scale economy was attained only through a single
monopoly. Indeed, given the state of technology of that time, the argument may have held. For in
point, the loaded copper transmission capabilities allowed only limited transport, namely one
voice channel per twisted pair. However, as we shall demonstrate, the underlying technology has
77 | P a g e
provided a dramatic change in the underlying system. Functions now provided by the network,
may be more efficiently provided by intelligent Customer Premise Equipment (CPE). The
question to be posed is; what is the role of the network, and how do we provide the dimensions
of creative freedom to allow these new roles to evolve? To effectively approach this problem, we
must first develop a canonical structure of a network.
Before continuing, we will define in a more structure fashion the paradigm, the world view, the
architecture and the technology base. First, Kuhn defines world view as:
" An entire constellation of beliefs, techniques, and so shared by the members of the community."
And he further defines the paradigm as, first:
"One sort of element in that constellation, the concrete puzzle-solutions which employed as
models or examples can replace explicit rules as a basis for the solution of the remaining puzzles
of normal science."
and then second as:
" Paradigms are what the members of the scientific community share and conversely a scientific
community consists of men who share a paradigm."
We however, take from these definitions and place them into the context that we have developed
in this section.
Definition: A paradigm is a unique and defining experiment or demonstration that in and of itself
crystallizes a concept as representative and descriptive of a broader class of similar ideas.
The classic paradigm was that of Watson and Crick in developing the structure of the DNA
molecule. The defining moment was best described in their paper in Nature that off-handedly
alludes to the DNA molecule having the properties of self-replication and thus containing the
genetic information for all life. This defining experiment then led to the massive changes in
Botany, Biology, Medicine and even computer science. The paradigm is the rite of initiation for
any believer in the new religion. The paradigm for Christianity was the resurrection of Jesus, and
that of Judaism was the pact of Abraham with the Lord.
To understand all other elements of deconstructionist thought, it is first necessary to find and
identify the defining moment, or the paradigm.
78 | P a g e
Definition: The world view is the collection of all beliefs that follow from a consistent
application of the paradigm in reflecting life and its existence.
The world view of the DNA paradigm is genetic engineering, of the Univac paradigm, is the
central processor, of the Church in the Middle Ages is salvation after death. The world view is
more than a single statement, it is a collection of beliefs. All the beliefs are predicated on the
unique and singular interpretation of the paradigm. Until Martin Luther, the Church in Rome was
in control. Luther, through the printing press and the availability of the bible to all challenged
and changed this. The new paradigm in this case was the printed bible in vernacular. The world
view was that of Protestantism.
All too often we see the world view in parts. Even if we see the world view in total we may fail
to see the paradigm. The religious organizations typically define and deify their paradigms. Jesus
at the mount, Mohammed and the angel, Joe Smith and the angel, May Baker Eddy and divine
revelation, and even the Buddha and his revelations. They are the paradigms. In religion it is
critical to have the paradigms unchallenged. They are thus constructed.
Definition: The technology base is defined as the set of all technologies available to demonstrate
the paradigm and to implement the representation of the world view.
We define technology in its broadest sense. We may have a paradigm of a god who has shown
themselves through the visions of a mushroom. The world view is that we must perform all of
our important acts under this influence. The technology allows us to analyze the mushroom and
determine based on our neuroscience knowledge that the effects are those of psilocybin, not of
our god. That will challenge the paradigm. Another is the ability to measure the age of the earth
and to carefully, through genetic analysis, determine the evolutionary pathways of all creatures.
This may challenge the paradigm of the creationists. A third, and current one is the technology to
determine that homosexuality is genetically based. This challenges the paradigm of choice in the
homosexual community and thus attacks a power base. Technology can thus either empower the
world view and express it or it can challenge and destroy the paradigm.
Definition: Architecture is defined as the conceptual embodiment of a world view, using the
commonly understood set of constructural elements, based upon the available set of
technologies.
Definition: The design is the current implementation of an architecture using the current
technology base.
Definition: The process of deconstruction is the process whereby the current observer, by using
the understanding of the technology base evolution, and by understanding and structuring the
79 | P a g e
current and prior architectures, can determine the base paradigm and thus infer all elements of
the world view of the implementor of the architecture.
We shall use this process in the development of many multimedia systems. The design of a
multimedia system must understand the paradigms of the users. This is inclusive of all of the
user sets, namely the food chain elements that we discussed before. To deconstruct, we must
understand and identify the paradigm. Without that we cannot hope to achieve the
converasationality that we seek.
Finally, we add the definition of an infrastructure, recapturing what we developed in the last
section.
Definition: An infrastructure is a shareable, common, enabling, enduring, resource, that has scale
in its design, and is sustainable by an existing market, and is the physical embodiment of and
underlying architecture.
2.7 QUESTIONS
We proceed with the effort of developing a deconstruction theory by developing a set of key
questions that ultimately relate to our understanding of object or entities. It will be through this
understanding that we come to grasp the underlying structures of multimedia.
1. What is an object?
An object is and entity in a multimedia environment that is actionable. It is a single or compound

collection of multimedia elements, combined from one or several users, that creates a
transaction. Actionable implies that the objects can be used by others or shared amongst others
and the sharing or use results in human actions or responses. The concept of actionable objects is
at the heart of multimedia communications.
2. Do Objects exits?
Objects exists through the transactions that they create. Does a conversation exist? Words are
created, sounds are made, and a consensus is achieved. Possibly the objects used in the
conversations are perceived differently by each of the participants, and possible the consensus is
perceived differently, even if it is placed in writing and agreed to by all. The true sign of the
objects existing is their ability to create a transaction at some point, namely a change element.
3. What role does the human play in the expression of the characteristics of an object?
80 | P a g e
Does a tumor such as a lymphoma exist? The patient presents with a symptom of a swelling in
the neck. The physician palpates and remembers the past 99% of the cases were lymphomas, by
definition, the pathologist reads the slide and sees the telltale cell pathologies. The object of the
lymphoma exists. The existence is in the consensus of the individuals. The ultimate existence is
in the life or death of the patient
4. How does a human deal with objects?
The human deals with the set of objects through their totality of senses. The object may be
visual, all too frequently what we are trained to respond to in today's culture, it may be a sound, a
touch, a smell, or any of the other sensual combinations. The human rationalizes the sense
objects and creates a communal sense of agreement with others with regard to a collection of
such objects. The human deals with multimedia communications objects in a conversational and
communal fashion. Unlike a book, or a video game, which is an internalized experience, a
multimedia communications based object is externalizable, and allows for the displaced
conversationality described earlier.
5. Why does a human deal with an object?
The human deals with the objects through their senses and with the interaction of other humans.
It is a consensus-conflict resolution process.
6. Why does a human react to the presence of an object?
The human reacts to the object by transacting an event. The multimedia objects are essentially
demands on the individuals senses that will not go unanswered. It is an active medium that
demands a reaction and response. Unlike television which may be internalized and has responses
that are displaced in their cultural context, multimedia communications is in the context of a real
time communal dialog.
7. "What is communications?"
Communications in the context of multimedia is a conversational transaction. It allows multiple

individuals, humans or otherwise to interact, converse, and through that conversation to transact.
Communications is the movement of information between individuals, movement that entails the
interaction of as many of the sense as is possible.
8. "What is conversation?"
81 | P a g e
Conversation is the sharing of senses, the establishment of consensus or the resolution of
conflict, with the end result be the transaction. A conversation requires two or more humans, a
common shared medium, co-temporaneous or otherwise, and a shared set of beliefs, that allow
for a common understanding of the signs used on the conversation. The semiotician would look
at the conversation in terms of the external signs. The deconstructionist would look at the
underlying pre-beliefs of the communications. The hermeneuticist would try to understand the
motives of the messengers in the conversation. In the case of multimedia communications we are
using signs that are externalities of ideas that we wish to transmit from one individual to another.
The end result of this transmission of ideas is a transaction, a change in the state of the receiver
or even the sender of the ideas.
Multimedia communications is about conversations. It is about using all of the constructs

available to communicate information that results in that desired or otherwise transaction. The
challenge is to determine what elements are necessary for the state change. The further challenge
is to determine if that change really occurred.
9. "How do humans interact with information?"
Humans interact with information by responding and creating a transaction. Information allows
the change of knowledge state in the individual. If I am provided with a new piece of
information, I change my understanding of something that I had not fully understood before. It
may clarify it or it may dramatically alter it. If there is no change, then we argue that there was
no information.
10. "What is a representation?"
A representation is an embodiment of a collection of information elements. In a multimedia

communications environment we are almost always dealing with representations and never the
actual entity. For example, in a radiological environment, we look at the MRI of the brain of a
patient with multiple sclerosis. We see in the brain the white spots that represent the
demyelination of the brain cells, the evidence that there is a sclerotic event. If we were the
surgeon, we would most likely see nothing, if the pathologist, we would see the demyelination
on a different scale by staining the cells and see the loss of myelin. All are representations of the
underlying disease process. The reality is as best defined a genetic and viral induced process that
is the humans immune system rejecting the self. This is the long winded way of defining the
entity via the process as compared to defining the entity via the representation. The
representation defines an entity that may be defined in no other way.
11. "Does there exist an abstraction that can be shared by two people that allows common
understanding?"
82 | P a g e
The essence of multimedia communications is the delivery of abstractions. An abstraction is a
representation of an entity delimited by the technological resources available in the multimedia
communications environment. A full representation may require all the senses, or a breath of the
senses that exceeds what is generally available to the current technology. The abstraction is the
mapping of the representation onto the technology base. Sharing abstractions may allow two or
more people to attain a common understanding if there is adequate information conveyed in the
abstraction. For example, two physicians may see a chest x ray and determine that the patient has
a carcinoma of the left lung. The spot is clear and non-disseminated, and based upon mutual
shared prior experience it is clear what the abstraction is saying. If one of the viewers is the
patient who is not medically trained, the abstraction would not convey such information. Thus
information exchange requires a shared abstraction as well as a shared experience base with that
set of abstractions. Expecting that abstractions per se convey information is not viable.
12. "Does a representation exist independent of the observer."
This is the dualism between forma and substance, between essence and existence. Is there such a
thing as a rose, or is there a set of observable that when clustered together reach the consensus in
most people that this is a rose. The body of plant Systematics faces this problem at all times. The
answer is that there is no such thing as an abstract representation. Information and determination,
the transaction, take place with a set of noisy observations that have been characterized and
categorized.
13. "Does a representation change in its meaning when observed by more than one observer."
This is the same as asking if consensus changes if there are more or less participants in the
process. Will a jury meet a different decision if there are different jurors. The answer is possibly.
Consensus is convergent but there is no abstract convergent point.
Why are these questions important and how do they relate to multimedia communications.
Multimedia communications consists of communicating objects that relate to the human in only
one fashion; through a transaction or through the empowerment to act. When designing a system,
we empower the users to act. They must do so through the use of objects, the manipulation of the
objects and the gaining of a consensus.
4.3 Answers?
The questions posed in the last section and the discussion concerning them clearly indicate that a
certain set of long standing issues are still at the heart at how we view ourselves as creatures. The
83 | P a g e
main driving issue has been the mid and body dualism. As we creatures that think with and act
with our minds, or are our bodies the totality of ourselves. To quote from Winograd and Flores:
"...mind-body dualism...rests on several taken-for-granted assumptions:
1. We are inhabitants of a real world made up of objects bearing properties. Our actions take
place in that world.
2. There are objective facts about the world that do not depend on the interpretation or even
presence of any person.
3. Perception is a process by which facts about the world are registered in our thoughts and
feelings.
4. Thoughts and intentions about action can somehow cause physical motion of our bodies."
This concept of the ideal form and the ideal an achievable entity is as old as Plato and Aristotle.
The concept of the ideal form, as a Platonist would state, is that there is a true idea of a daylily. It
is an abstraction that is the daylily, and what we see as humans is a mere shadow of its true form.
To the multimedia communications, we then ask how does a Platonist communicate, namely,
does he try to use the abstraction that closely matches the form? Copleston speaks on this with
regard to Plato:
" I would point out that the essence of Plato's doctrine of Forms and Ideas is simply this: that the
universal concept is not an abstract form devoid of objective content or references, but that to
each true universal concept there corresponds an objective reality."
Continuing he states further:
"In the Republic it is assumed that whatever a plurality of individuals have a common name, they
have also a corresponding idea of form. This is the universal, the common nature or quality
which is grasped in the concept."
It is the attempt to describe the "nature" or essence of things and to use this as a means to
communicate that is the basis of many of our problems in design. An example is the compression
of speech or video. We compress to avoid the need for more bandwidth. We compress also
because we believe that by doing so we get to the essence of it. We do so in a Shannoesque
fashion, assuming that there is an essence of bits, minimal as they may be. This extension is best
described by Popper:
84 | P a g e
" I use the name methodological essentialism to characterize the view, held by Plato and many of
his followers, that the task of pure knowledge or "science" to discover and to describe the true
nature of things; their hidden reality or essence. ...All these methodological essentialists also
agreed with Plato in holding that these essences may be discovered and discerned with the help
of intellectual intuition. A description of the essence of the thing they called the "essence"."
An extreme position to this essence approach is the positivist approach expressed by Ayer when
describing the early work of Wittgenstein.
"..the main theses of the Tractatus can be easily summarized. The world is said to be totally of
facts which themselves consist in the existence of what are called.. atomic facts.. or states of
affairs. The states of affairs consist of simple objects, each of which can be named. The names
can be significantly combined in ways that express elementary propositions. Each proposition is
logically independent of all its fellows. They are all positive and each of them depicts a possible
state of affair which constitutes its sense....The fact that they are logically independent means
that in order to give a complete account of reality one has to say which of them is true or false."
The development of multimedia is the development of new metaphors. MacCormac best

describes this change that metaphor can take:
" Metaphor can be described as a process in two senses: (1) as a cognitive process by which new
concepts are expressed and suggested, and (2) as a cultural process by which language itself
changes...epiphors are metaphors that express more than they suggest..diaphors suggest more
than they express."
He goes on to state:
"Generations of students who have passed through introductory philosophy courses in colleges
and universities have come to believe in the division between the mind and nature. The rise of
cognitive psychology in opposition to behaviorism, which denied the existence of the mind, fins
comfort in the philosophical efforts to build a foundation for knowledge. The account that I have
presented of metaphor as a cognitive process presumes the existence of the mind existing as a
deeper level of explanation that of semantics and surface language."
The essence of the Heidegger philosophy as relates to multimedia design has been best described
by Winograd and Flores:
"We...present...a...discussion of Heidegger's philosophy,...
(1) Our implicit beliefs and assumptions cannot be all made explicit.
85 | P a g e
(2) Practical understanding is more fundamental than detached theoretical understanding.
(3) We do not relate to things primarily through having representations of them.
(4) Meaning is fundamentally social and cannot be reduced to the meaning-giving activity of
individual subjects."
The final element of Heidegger's approach is the breaking down effort of providing information
in a way in which it is broken down or handled by the user.
"... Heidegger's ...insistence that objects and properties are not inherent in the world, but arise
only in an event of breaking-down in which they become present-at-hand...In sum, Heidegger
insists that it is meaningless to talk about the existence of objects and their properties in the
absence of concernful activity with its potential for breaking-down."
The latter comment on Heidegger is the essence of multimedia communications. The breaking
down if the basis of a transaction, of a change in state of the human or humans in the
conversation. We must then take into account the impact of this new medium. It is an impact
with many dimensions of consequences. To quote McLuhan:
"The personal and social consequences of any new medium result from the new scale that is
introduced into our affairs by each extension of ourselves or by any new technology."
And to further the quote:
"The message of any medium or technology is the change of scale or pace or pattern that it
introduces with human affairs."
The multimedia revolution is on the scale of all other revolutions. It will be a revolution if and
only if it does to the multimedia word what the printing press did to the written word. To quote
McLuhan:
"The French Revolution, as per de Tocqueville, was a result of the homogenizing nature of the
printed word."
Homogenizing means making it accessible to all. Making it accessible means making it

actionable, and actionable at a distance. The actionable at a distance, and the ability to have the
86 | P a g e
throwness in the medium and the integration with the message is what will make for the
revolution.
2.8 HERMENEUTICS
Deconstruction is the process whereby the current reader attempts to place themselves in the
context of the writer, both used in generic terms, and determine what the message was that was
meant to be sent, relative to the context of it being sent. It in many ways is a Bayesian analysis of
a human communications process. Hermeneutics is complementary to this effort. Hermeneutics,
named after the god Hermes, is in essence the attempt to understand the environment of the
singer and writer of the songs, but to sing them as clearly and faithfully as one can.
2.8.1 Hermeneutic Principles
Hermeneutics is originally imbedded in the interpretation of texts. In our analysis the "text" is the
broadened entity of the multimedia environment. The "text" must be expended into the context of
the Messenger as carrier of the Message. Thus the need for a hermeneutic understanding is to
focus on the messenger and what does the messenger bring to us about the message. In very
practical terms, therefore, the same hermeneutic arguments that allowed us to address how to
interpret the bible, are and will be at the heart of how do we interpret an x-ray and blood smear
in a multimedia environment. The question posed is that does the multimedia environment land
through the changed medium ad different message since it is communicated by a different
messenger. Or, do we try to ensure that he messenger is kept intact. Another analogy is does the
bible change when one understands it from televangelists rather than reading it. Or historically,
does the bible change when each person reads it as compared to having it preached from the
pulpit. The latter change led to the reformation and the end of the hegemony of Catholicism in
Europe.
The hermeneutical school is a contrast to the positivists who argue that one can obtain objective
knowledge. To the hermeneutic student, all knowledge is "interpretation". It is in this noisy
channel that we try to obtain information. It is thus upon this noisy information that we
ultimately act. The multimedia architect must take this into account in the designs of their
systems. The channel is inaccurate and furthermore the messenger may actually be devious. The
try of hermeneutics attempts to address the issue of devious messengers in the context of texts.
We argue that the same questions and approaches are essential for the development of
"multimedia text" messages.
The issue relating to texts is developed in Warnke. She states the problem in a historical context:
"Questions of interpretation had been raised earlier, in particular in the Reformations Challenge
to the catholic reign of the bible. Did an understanding of Scripture require a prior acceptance of
87 | P a g e
the precepts of the Catholic faith or could it be understood on its own?...Schleiermacher
significantly expanded the scope(asking) how many could be comprehended, what methods
would permit an objective understandings of texts and utterances of any kind...Dilthey even
asked broader questions: what were the methods that would permit an objective reading of any
kind, including actions, social practices, norms and values? How could the understanding of
meaning be raised to the same methodological clarity characterized in the natural sciences? How
could it find as solid a basis for methodological progress?"
Gadamer is the most recent and articulate espouser of the hermeneutic school. He has evolved
and matured the hermeneutic approach from one of literal translation to exposition. To quote
Warnke describing the evolution in hermeneutics in Gadamer;
"The Bible is assumed to have a normative authority for everyone and the task of the
hermeneutic understanding is therefore simply to help transmit the content of its normative
claims."
Simply put, Hermeneutic in this context states that the deconstuctionist approach may be use in a
relative setting, but in the context of a normative setting such as the law of God, we us either
hermeneutic approach to seek the "truth" or normative facts. The difference is best stated in
terms of selecting a justice of the Supreme Court. Judge Bork is the classic hermeneutic seeking
the letter of the constitution. Justice Douglas is the decosntructioninst trying to take Madison's
words and placing them in a current time frame.
Gadamer takes the hermeneutic goal of positivism and objective answers and introduces the
subjective. This is characterized again by Warnke:
"Hermeneutics, as Gadamer conceives of it, then, is no longer to be seen as a discourse on

methods of the "objective" understanding as it was for the hermeneutic tradition of
Schleirmacher and Dilthey. It no longer seeks to formulate a set of interpretative rules; rather, in
referring to his analysis as "philosophical hermeneutics", Gadamer turns to an account of the
possibility of understanding in general, conditions that in his view undermine faith in the ideas of
both method and objectivity. Understanding is therefore rooted in prejudice and the way in
which we understand it is thoroughly conditioned by the past of by what Gadamer calls "effective
history"."
From our perspective, as designers of multimedia systems, we have a cultural environment that
we are working in. We have an environment with a history, a past, a culture, and a noisiness that
makes event objective transmission of information a transformation of information. The
hermeneutic channel challenge is to model the channel, to deconstruct its structure, to generate
the optimal processing filters for the complex messages that are to be transmitted.
88 | P a g e
From a historical perspective, the hermeneutic problem for the multimedia designer is the same
as the random noise problem was for the designer of signal detection systems of the 1940's and
1950's. As we had indicated before, it was the work of Shannon who determined that information
was essential the elimination of uncertainty. Shannon's teacher, and in many ways mentor, was
Norbert Wiener. It was to Wiener, who had both mathematical and philosophical raining, that the
development of the concepts of the detection of signals in noise is credited. Wiener conceived of
the use of the correlation and auto correlation functions. He introduced the history of an random
process in a structured for that has led to information theory and the processing of signals in
computers and communications. It is the same construct that we are trying to develop that is of
utmost importance. The mathematical theory may not be in place, but the philosophical
constructs to develop them must first be worked through.
Warnke goes on in terms of hermeneutic development:
"Hermeneutics thus has a largely pedagogical task: it is supposed to exhibit the truth that inheres
in a given claim so that its audience can understand and learn from it. As hermeneutics develops,
however, attention is redirected from the understanding of truth context of a text and towards the
understanding of the intentions. The aim of understanding is no longer seen as knowledge of die
Sache- a substantive knowledge of claims to truth or normative authority. It is seen rather as
insight into the historical and biographical circumstances behind their expression. Understanding
becomes genetic: what were the conditions under which the agents acted, spoke or wrote as they
did?"
This leads to a focus on the circumstances qua written word. Deconstruction focuses on the
meaning qua circumstances. We must enter into the hermeneutic thought process from the
perspective of multimedia communications because the displacements of Gadamer and
Haberamus are physical and temporal, but the displacements of multimedia communications are
electronic and cultural. The nature of displaced understanding is the same in both cases. The
issues, we argue, are also isomorphic.
The complementarity of approaches can be related to the complementarity of ideas in cultural

context. For example, Bloom recounts the issue in the context of the French;
" Descartes and Pascal are national authors, and they tell the French people what their
alternatives are, and afford a peculiar and powerful perspective on life's perennial problems...On
my last trip to France I heard a waiter call one of his fellow waiters a "Cartesian"...Descartes and
Pascal represent a choice between reason and revelation, science and piety."
89 | P a g e
These two authors are also choices between two types of certainty; the divine and the mind. In
reality, as we have seen with Gadamer, certainty of any kind has its limits. The positivist school
still argues for the existence of absolute certainty. The hermeneutic school of Gadamer eschews
such certainty. At best we can interpret. At worst, the interpretation is a reflection of our own
past, history, biases and intents. Even in medicine, the process of diagnosis is one of noisy
interpretation. Certainty may exist in a pathology slide that portrays without doubt a malignancy.
The outcome of that diagnosis may still have some uncertainty.
Namely, there is a philosophical underpinning in our general line of communications. It is this

basis that will lead to understanding or cacophony. Gadamer goes on to define his focus in
hermeneutics as follows;
" The task of philosophical hermeneutics, therefore, is ontological rather than methodological. It
seeks to throw light on the fundamental conditions that underlie the phenomenon of
understanding in all of its modes."
Gadamer further stresses the importance of language in this process:
" Language is the fundamental mode of operation of our being-in-the-world and the all-
embracing form of the constitution of the world."
The issue of language being the form of history, the carrier of the formation, and the medium of
the message, and thus being the message itself, has implications to the hermenuticist. One of
these is the change in the nature of the media for the transmission of the e message. Weizenbaum
notes:
" The computer has thus begun to be an instrument for the destruction of history. For when a
society "legitimates" only those data that are "in the standard format" and "that can easily told to
the machine" then history, memory itself, is annihilated. The New York Times has already begun
to build a database of current events...from which historians will make inferences as to what
really happened."
The conversion of words in the twelfth century, with the rediscover of the Greeks and their
thoughts, which arguable led to the enlightenment of the twelfth century itself, was a result of the
translation, and not simple transliteration of the Greek texts. As Illiach and Sanders remark:
" The Greek work was not to be turned into Latin verbum pro verbo. Instead, the meaning was to
be detached from the words of one language and made to reappear in another; content, stripped
of its form, was to be preserved. Theories about translation changed very little - translation was
described as an attempt to divulge the secrets of one language into another- until the
90 | P a g e
hermeneutics of the 1950s. Only then did the study of translation as applied linguistic theory
become separated from literary theory."
Ironically, the twelfth century Latin of the early universities, such as those in Paris, was a Latin
ready for the expansion of the new technology of Guttenberg. It was now a language whose form
was prepared for text. Again Illich and Sanders remark;
" Division into words first came into common use in the seventh century. It happened at the
northern frontiers of the known world, where Celtic "ignoramuses" had to prepare for the
priesthood and needed to be taught Latin. Division of words was thus introduced as a means of
teaching Latin to barbarians as a foreign language."
The very structure of language had made a transition into a form that would allow it to be further
tempered by a technology and thus be transformed into the new medium. The bilateral change of
the understanding of the environment of the author and the tempering of the environment of the
reader then leads to the full hermeneutic context. As Winograd and Flores remark:
" .. Gadamer takes the act of interpretation as primary, understanding it as an interaction between
horizon provided by the text and the horizon that the interpreter brings to it. Gadamer insists that
every reading or hearing of a text constitutes an act of giving meaning to it through
interpretation."
Thus the hermeneutics of Gadamer is evolutionary from the revelation of the underlying eternal
truth, to the evolving interpretation.
2.8.2 Hermeneutic Methodology Applied
In this section we take the philosophical theory developed in the prior section and address it to
the problems of the multimedia methodology. We develop the design principles in hermeneutic
observations. The process is to state the observation and then to reflect on the appropriate design
principle.
1. What we observe is a reflection of what is there filtered through the understanding of

what we think should be there.
2. Conversations have history, and the history is often unknown, and if known may not be
aware to the conversant.
3. Absolute "truth" does not necessarily exist in a conversation. Consensus may converge
but convergence is not to truth. The convergence of consensus is not consistent.
91 | P a g e
4. Tradition, authority, and history are integral elements in filtering understanding. The
issue of authority is an integration of visions of the presenter and the information presented.
Choice of words and language to create a "text" are the basis of power in the relationship.
5. Communications is a sociological interpretive process that seeks to attain consensus in

resolving an ambiguity. Actions are the result of that consensus becoming an agreed to common
state amongst the community in the process.
2.8.3 Communications at the Conversational Layer: Hermeneutics Applied
Having established the design rules and provided the understanding of the hermeneutic elements,
we can now take these and place them in the context of multimedia communications. The issue is
one of establishing conversationality. Coversationality is as we have defined the ability of one or
more humans to enter into a dialog with a system and themselves relating to elements of
information in the system adequate enough to eventually create a transaction.
The OSI layered communications architecture has evolved to manage and support the distributed
communications environment across error prone communications channels. It is presented in
detail in either Tannenbaum or Stallings. A great deal of effort has been spent on developing and
implementing protocols to support these channel requirements. Layer 7 provides for the
applications interface and generally support such applications as file, mail and directory. The
requirements of a multimedia environment are best met by focusing on layer 5, the session layer
whose overall function is to ensure the end to end integrity of the applications that are being
supported.
Some authors (See Couloris and Dollimore or Mullender) indicate that the session function is
merely to support virtual connections between pairs of processes. Mullender specifically deals
with the session function in the context of the inter-process communications (IPC). In the context
of the multimedia object requirements of the previous section, we can further extend the concept
of the session service to provide for IPC functionality at the applications layer and specifically
with regards to multimedia applications and their imbedded objects.
The services provided by the session layer fall into four categories:
1. Dialog Management: This function provides all of the users with the ability to control, on
a local basis as well as global basis, the overall interaction in the session. Specifically, dialog
management determines the protocol of who talks when and how this control of talking is passed
from one user to another.
2. Activity Management: An activity can be defined as the totality of sequences of events

that may be within a session or may encompass several sessions. From the applications
92 | P a g e
perspective, the application can define a sequence of events called an activity and the session
service will ensure that it will monitor and report back if the activity is completed or if it has
been aborted that such is the fact.
3. For example, in a medical application, we can define an activity called "diagnosis" and it
may consist of a multiple set of session between several consulting physicians. We define a
beginning of the activity when the patient arrives for the first visit and the end when the primary
physician writes the diagnosis. The session service will be responsible for ensuring that all
patients have a "diagnosis".
4. Synchronization: We have seen that at the heart of a multimedia system is a multimedia

data object. Each of the objects has its own synchronization or timing requirements and more
importantly, a compound object has the orchestration requirement. The session service of
synchronization must then ensure that the end to end timing between users and objects is
maintained throughout.
5. Event Management: The monitoring of performance, isolation of problems, and

restoration of service is a key element of the session service. Full end to end network
management requires not only the management of transport and subnetwork, but requires that
across all seven OSI layers, that overall end to and management be maintained (See McGarty
and Ball).
The servers are conceptually at a level above the transport level. We typically view the transport
servers as communicating distributed processes that are locally resident in each of the
transmitting entities. This then begs the question as to where does one place the session servers.
Are they local and fully distributed, can they be centralized, and if so what is their relationship to
the Transport servers. Before answering these questions, let us first review how the session
services are accessed and how they are communicated.
Session services are accessed by the higher layer protocols by invoking session service
primitives. These primitives can invoke a dialog function such as Token_Give. The application
may make the call to the S_SAP and this request may be answered. There are typically four steps
in such a request, and these are listed in Stallings who shows that the requests are made of the
session server by entity one and are responded to by entity two. The model does no however say
where the session server is nor even if it is a single centralized server, a shared distributed server,
or a fully distributed server per entity design. We shall discuss some of the advantages of these
architectural advantages as we develop the synchronization service.
5.3.1 Dialogue Management
93 | P a g e
Dialog management concerns the control of the end user session interaction. Specifically, who
has permission to speak and when, who can pass the control and how is that implemented. In this
section we shall describe the environment for the dialogue management function and develop
several possible options for implementing this function.
Dialog management requires that each of the virtual users have a token or have access to a baton
in order to seize control of the session. In the course of a typical session, the two virtual users fist
establish the initial subsession that becomes the first part of the session. The addition or binding
of other virtual users through subsessions to the session allows for the growth of the session. The
baton or token may be a visible entity that is handed from one to the other or it may be hidden in
the construct of the applications.
Consider the session level service called dialogue. The service can be implemented in four
possible schemes. These schemes are:
(1) Hierarchical: In this scheme there is a single leader to the session and the leader starts as the
creator of the session. The baton to control the session can be passed upon request from one user
to another, while full control remains with the session leader. The session leader may relinquish
control to another user upon request and only after the leader decides to do so. The leader passes
the baton from users to user based upon a first come first serve basis. It is assumed that each
users may issue a request to receive the baton, and that any requests that clash in time are
rejected and the user must retransmit. There transmit protocol uses a random delay to reduce the
probability of repeated clashing. The leader always acknowledges the receipt of the request as
well as a measure of the delay expected until the baton is passed.
(2) Round Robin: In this scheme, the baton is passed sequentially from one user to another. Each
user may hold the baton for up to Tbat sec and then must pass the baton. When the baton is held,
this user controls the dialogue in the session.
(3) Priority: In this case, all of the users have a priority level defined as Pk(t), where k is the user
number and t is the time. We let the priority be;
Pk(t) = Rk(t) + Tk(t) + Dk(t)
Here R is the rank of the k th user, T is the time since the last transmission and D is the data in
the buffer. We assume that some appropriate normalization has occurred with this measure.
94 | P a g e
Every Tcheck seconds, each users, in sequence sends out a small pulse to all other users, on a
broadcast basis, and tells them their current priority. Each user calculates the difference between
theirs and all the others. User k calculates a threshold number, TRk, which is;
TRk = max |Pk(T)-Pj(T)|
If TRk > 0, then user k transmits its packets for Tsend seconds.
(4) Random Access: Each user has a control buffer that indicates who has control of the session,
namely who has the baton. The session is broken up into segment Tsess in length, with Treq
seconds being relegated to a baton ownership selection period and Tsess-Treq being left for the
session operation. During Treq , all of the users transmit a request packet that is captured by all
of the other users buffers. Treq is broken into two parts, Tsend and Teval. These requests are
broadcast in Tsend.
Now after the sent messages are received, one of two things can happen, the message is received
or it collides with another message and is garbled. If the message is garbled, the buffer is not
loaded and is left empty. If it is filled, then each buffer during Teval sequentially broadcasts its
contents and all of the users listen to the broadcast and record the counts, Nk where Nk is the
number of votes for user k in that call period.
The choice of baton control is then;
Choose user k if Nk = max j | Nj |
else restart Treq again.
5.3.2 Activity Management
Activity management looks at the session as an ongoing activity that users may come and go to.
This services provides an ability to easily add, delete and terminate the entire session.
An activity in the terms of the session is a total bounded event that can be compartmentalized in
such a way that other events may be locked in suspension until that event is complete. Activity
management is in the session layer a function similar to transaction management in a transaction
processing system. It allows for the definition of demarcation points that permit suspension of
activities in other areas until the activity managed transaction is complete. Activity management
can also be developed to manage a set of events that can be combined into a single compound
event.
95 | P a g e
There are several characteristics that are part of activity management:
1. Activity Definition: This allows for the defining of an activity as composed of several
dialogue. It allows for the defining of the activity as a key element of a single session or even to
expand over several sessions. Activity definition is the process of informing the session server of
the beginning and end parts of an activity and in providing the session server with an identifiable
name for the activity.
2. Activity Integrity Management: Activities are integral elements of action that cannot be
segmented. The activity management system must ensure that once an activity is defined and
initiated, hat no other activity that could interfere with the existing one is allowed to function.
3. Activity Isolation: The ability to provide integrity is one part of managing the activity.
Another is the ability to isolate the activity from all others in the session. An activity must be
uniquely separable from all other activities, and this separation in terms of all of its elements
must be maintained throughout its process.
4. Activity Destruction: All activities must be destroyed at some point. This is a standard
characterization.
There are several sets of activities that are definable in a multimedia environment. These are as
follows:
1. Compound Multimedia Object Transfer
2. SubSession Management
3. Dialog Control
The algorithms to perform the activity management functions are developable consistent with the
OSI standards. There are no significant special development necessary.
5.3.3 Synchronization Management
Synchronization is a session service that ensures that the overall temporal, spatial and logical
structure of multimedia objects are retained. In this case we have a source generating a set of
Voice (VO), video (VI), and Image (IM) data objects that are part of a session. These objects are
simple objects that combined together form a compound multimedia object. The object is part of
an overall application process that is communicating with other processes at other locations.
96 | P a g e
These locations are now to receive this compound object as show with the internal timing
retained intact and the absolute offset timing as shown for each of the other two users.
In this example, the synchronization function provided by the session server to the applications
processes at the separate locations is to ensure both the relative and absolute timing of the
objects. The location of the functionality can be centralized or distributed. Let us first see what
the overall timing problem is. Consider a simple SMO synchronization problem. The network
than transmits the packets and they arrive either in order or out of order at the second point. The
session server must then ensure that there is a mechanism for the proper reordering of the packets
at the receiving end of the transmission.
Let us consider what can happen in this simple example.
First, if the BMO of the SMO is very lengthy, then as we packetize the message, we must
reassemble it in sequence for presentation. Let us assume that the BMO is an image of 100
Mbits. Then let us assume that the packet network has a packet delay that will be low if there is
no traffic and grows as traffic increases. Now let us assume that the network provides 500 bit
packets transmitting at 50 Mbps.
Second, let us note that there are 200,000 packets necessary to transmit the data. Each packet
takes 10 microseconds to transmit. If we assume that there is a load delay of 5 microseconds per
packet, then the total transmit time goes from 2 to 3 seconds.
We can also do the same with a compound object. In this case, we take the CMO and note that it
is composed of SMOs. The SMOs must then be time interleaved over the transmission path to
ensure their relative timing. It is the function of the session service to do this. The application
merely passes the CMO and its header information as a request to the session server to ensure the
relative timing is maintained.
Here we have a CMO entering the network, knowing that the session server at Server 1 must not
only do the appropriate interleaving but it must also communicate with the other servers (in this
case K and N) to ensure that de-interleaving is accomplished. We show the session servers
communicating with the network through the T_SAP and that in turn takes care of the
packetizing. However, we also show that the session server, 1 and N, communicate in an out of
band fashion, using some inter process communications (IPC) scheme, to ensure that the relative
actions are all synchronized amongst each other. We can now envision how the architecture for
this can be accomplished. There are two schemes:
Centralized: It assumes that each application (A) has a local client (CL). The application
communicates with the local client (CL) to request the session service. The session server is
97 | P a g e
centrally located and communicates with the application locally by means of a client at each
location. This is a fully configured client server architecture and can employ many existing
techniques for distributed processing (See Mullender or Coulouris et al).
Distributed: In this case we have a set of applications, and cluster several applications per session
server. We again user local clients to communicate between the session server and the
applications. The clients then provide local clusters of communications and the session servers
allow for faster response and better cost efficiency. However, we have introduced a demand for a
fully distributed environment for the session managers to work in a distributed operating system
environment. As a further extreme, we could eliminate the clients altogether by attaching a
session server per applications and allow for the distributed processing on a full scale.
The major functions of the session server in its synch mode are:
To bind together simple objects into compound objects as requested by the application.
To provide intra object synchronization to ensure that all timing within each object is met.
To orchestrate amongst objects to provide inter object timing.
To minimize delay, slippage, between simple objects.
To minimize delay, latency, between different users.
To effect these requirements, we have developed and implemented a scheme that is based on a
paradigm of the phased locked loop found in communications (See McGarty and Treves,
McGarty). Here we have a distributed session server architecture receiving a CMO from an
application. The session server passes the message over several paths to multiple users. On a
reverse path, each server passes information on the relative and absolute timing of the CMO as it
is received using the session services primitives found in the OSI model. Generally for
segmented BMOs this is a simple problem but with streamed BMOs this becomes a real time
synchronization problem.
5.3.4 Event Management
Event Management deals with the overall end to end management of the session. It is more
typically viewed as a higher level network management tool for multimedia communications. In
the current sate this service is merely a reporting mechanism. Although ISO has expanded the
network management functionality of the seven layers, most of the functionality is still that of
98 | P a g e
event reporting. In this section we discuss how that can be expended for the multimedia
environment.
Event management at the session layer provides for the in band signaling of the performance of
the various elements along the route in the session path as well as reporting on the status of the
session server and the session clients. We note that each entity in the session path, which is
limited to all involved clients and all involved servers provide in band information on the status
of the session. In particular the in band elements report on the following:
Queue size at each client and server. The queue size can be determined on an element by element
basis.
Element transit and waiting time. For each element involved in a session, the time it takes to
transit the entire block as well as the time that the block has been in transit can be provided.
Session synchronization errors can be reported in this data slot. These errors can be compared to
lower level errors and thus can be used as part of the overall network management schema.
The structure of the event management system has been effectively demonstrated. It is
represented as a header imbedded in the transit of every data block. We can generate specific
event management blocks that are also event driven and not data transit driven. These are
generated by direct transmission of such blocks as overhead devoid of data content.
2.9 SEMIOTICS
The theory of semiotics is the theory of signs. Umberto Eco, academician, leader in the field of
semiotics, and author, has written several novels, one being the "Sign of the Rose". The novel is
about the fourteenth century and about a murder in a monastery. The hero uses the signs that are
left behind, the clues in common parlance, to play two roles. The first is the common role of the
clues to the murder mystery and the second as the signs to the change of the old guard to the
new. It is at the steps of the Renaissance and the signs of most importance are the books of old
knowledge. Thus Eco combines all elements of the deconstructionist, with those of the
hermenuticist and those of the semioticist.
2.9.1 A Definition
Eco defines semiosis as;
" Semiosis is the process by which empirical subjects communicate, communications processes
being made possible by the organization of significant systems. Empirical subjects, from a
semiotic point of view, can only be defined and isolated as manifestations of this double
99 | P a g e
(systematic and processual) aspect of semiosis. ...Semiotics treats subjects of semiosic acts in the
same way: either they can be defined in terms of semiotic structures or-from this point of view-
they do not exist at all."
The elements of semiosis comprises signals and signs. These Eco defines as:
" We are now in a position to recognize the difference between a signal and a sign...as sign may
be an expression system ordered to a content, but could also be a physical system without
any...purpose. A signal can be a stimulus that does not mean anything...a sign is always an
element of an expression plane... correlated to elements of a content plane."
Specifically, Eco combines these as:
"Semiotics is mainly concerned with signs as social forces."
Semiosis is based upon the extension of Pierces concept of pragmatism. The Pierce view of
pragmatism is presented by Copleston;
" Pragmatism, as Pierce conceives it,...is a method of reflection having for its purpose to render
ideas clear. ... Pierce divides logic into three main parts, the first of which is speculative
grammar...concerned with the ...meaningfulness of signs. A sign...stands for an object to
someone whom it arouses a more developed sign..the relation of significance, or the semiotic
function of signs, is for Pierce...a relation between sign, object and interpretant."
The issue of semiosis as a carrier of information is also embedded in the issue of a code. Eco
defines this as follows:
"..a code establishes the correlation of an expression plane with a content plane...a sign-function
establishes the correlation of an abstract element of the expression system with an abstract
element of the content system...a code establishes general types.. producing the rule which
generates concrete tokens...both ...represent ...the semiotic correlation and with which semiotics
is not concerned.."
The most descriptive distinction of what semiosis does in comparison to hermeneutics is best
described by Eco:
"In order to understand the history of Christian theology, it is not necessary to know whether a
specific actual phenomenon corresponds to the word, transubstantiation, ... it is necessary to
know what cultural unit ... corresponds to the content of that word.
100 | P a g e
The semiotic object of a semantics is the content, not the referent, and the content has to be
defined as a cultural unit.."
The totality of semiosis is its ability to combine its constructs with all elements of signals and
signs.
"...semiotics has been provided with a paramount subject matter, semiosis. Semiosis is the
process by which empirical subjects communicate, communications processes being made
possible by the organization of significant systems. Empirical subjects, from a semiotic point of
view, can only be defined and isolated as manifestations of this double... aspect of semiosis. ...
Semiotics treats subjects of semiosic acts...either they can be defined in terms of semiotic
structures of-- from this point of view -- they do not exist at all."
We can now develop some of the basis theory of semiotics as presented by Eco. We start with a
set of definitions and then discuss some of the implications.
Expression Plane: This is the plane of understanding in which one creates meaning through and
abstract expression. In physics, for example the expression plane may state the following:
This expression is the expression for the propagate of light or any other electromagnetic wave in
free space. Another example is the composition of a gene;
*ATTGTAAGCCGGATTTTC*
This is a set of nucleotides that imply their complements on the opposite side of the DNA pair.
This sequence means a great deal to a molecular biologist. They may imply blue eyes, green
wings, or red flowers. Both of these items are in the expression plane.
Content Plane: The plane in which one creates meaning through concrete expressions expressed
in a continuum of actions in the physical world. Simply put, it is everything that you can think
about it..
Let us take the propagation equation. This may be the expression of a rainbow. The content plane
is the rainbow and the full panoply of what it evokes. It is macro and micro in its expression. It is
the emotion that it may evoke leading to poetry, the grandeur it mat evoke in nature and
whatever else may do.
In the genetic case it is the fullness of the expression of the gene and its characteristics. It is the
expression of what green eyes may have and the human emotions that they may evoke.
101 | P a g e
In the case of multimedia, we design and implement our systems in the expression plane. We
express the in the content plane to ourselves as persons. The design problem will then be to
connect the expression and content planes effectively.
Sign: An element of the expression plane correlated to an element of the content plane. It is a
correlation recognized by the human.
Signal: A pertinent unit of a system that may be an expression system ordered to a content, or
could be a physical system with no semiotic purpose. Units of transmission which can be
computed quantitatively irrespective of their respective meaning.
Sign-Function: A relationship between system on E/C Planes, "Expression/Content".
Token: Types are generated by a code.
Token-Sign: Relation between units on E/C planes.
Type: Code generates types producing a rule which generates concrete tokens.
Code: Establishes general types, therefore producing the rule that generates tokens.
Connotative Relationships: A concatenation of the E/C plane such as shown below;
This example demonstrates the concatenation possible between the Expression and the Content
and how content may be concatenated upward itself. Eco makes this example one step further
when he states:
"Semiotics is mainly concerned with signs as social forces"
2.9.2 Semiotic Application
As we have defined the semiotic approach it deals with the signs or externalities of the process of
communications. In the multimedia world these externalities are the elements themselves, the
data objects, the storage objects and the actual human interaction with devices. All are signs.
Signs are means by which we relate from one plane upward to another. Signs in the multimedia
world focus on relating from the image and the test and the voice segment to the determination
of the diseased state. The design problem from the multimedia semiotician then is to go from
expression to content. The design problem for the seomiotic architect is the opposite; from the
content to the expression plane. How, does the semiotic multimedia architect ask, does one
102 | P a g e
ensure that the expression of the multimedia system contain adequate elements to reflect the
process in reverse and ensure the commonalty of expression by the richness of the expressionist?
2.9.3 The Multimedia Database; A Semiotic Example
We shall use the multimedia design problem of the database as a starting point in the application
of semiotic theory to the design of the system from an architectural perspective.
In a more standard computer communications environment, the data objects have significant
structure and they are frequently integrated into a system wide data base management system
that ensures the overall integrity of the data structures. In a multimedia environment, the data
elements are more complex, taking the form of video, voice, text, images and may be real time in
nature or can be gathered from a stored environment. More importantly, the separate data objects
may combined into more complex forms so that the users may want to create new objects by
concatenating several simpler objects into a complex whole. Thus we can conceive of a set of
three objects composed of an image, a voice annotation and a pointer motion annotating the
voice annotation. The combination of all three of these can also be viewed as a single identifiable
multimedia object.
Before commencing on the issues of communications, it is necessary to understand the data

objects that are to be communicated. We can consider a multimedia data object to be composed
of several related multimedia data objects which are a voice segment, an image and a pointer
movement (e.g. mouse movement). As we have just described, these can be combined into a
more complex object. We call the initial objects Simple Multimedia Objects (SMOs) and the
combination of several a Compound Multimedia Object (CMO). In general a multimedia
communications process involves one or multiple SMOs and possibly several CMOs.
The SMO contains two headers that are to be defined and a long data sting. The data string we
call a Basic Multimedia Object (BMO). There may be two types of BMOs. The first type we call
a segmented BMO or SG:BMO. It has a definite length in data bits and may result from either a
stored data record or from a generated record that has a natural data length such as a single image
screen or text record.
The second type of BMO is a streamed BMO, ST:BMO. This BMO has an a priori undetermined
duration. Thus it may be a real time voice or video segment.
A simple multimedia object, SMO, is a BMO with two additional fields; a Synchronization field
(Synch) and a Decomposition field (Decomp). The Synch field details the inherent internal
timing information relative to the BMO. For example it may contain the information on the
sample rate, the sample density and the other internal temporal structure of the object. It will be a
useful field in the overall end to end timing in the network.
103 | P a g e
The second field is called the Decomp field and it is used to characterize the logical and spatial
structure of the data object. Thus it may contain the information on a text object as to where the
paragraphs, sentences, or words are, or in an image object, where the parts of the image are
located in the data field.
These fields are part of an overall architecture requirement finds it necessary to provide an "out-
of-band" signaling scheme for the identification of object structure. The object structure is
abstracted from the object itself and becomes an input element to the overall communications
environment. Other schemes use in-band signaling which imbeds the signal information with the
object in the data stream. This is generally an unacceptable approach for this type of
environment.
When we combine these objects together we can create a compound multimedia object. A CMO
has two headers, the Orchestration header and the Concatenation header. The Orchestration
header describes the temporal relationship between the SMOs and ensures that they are not only
individually synchronized but also they are jointly orchestrated. The orchestration concept has
also been introduced by Nicolaou. In this paper we further extend the orchestration function
beyond that of Nicolaou. The concatenation function provides a description of the logical and
spatial relationships amongst the SMOs.
We can also expand the concept of a CMO as a data construct that is created and managed by
multiple users at multiple locations. In this construct we have demonstrated that N users can
create a CMO by entering multiple SMOs into the overall CMO structure.
The objectives of the communications system are thus focused on meeting the interaction
between users who are communicating with CMOs. Specifically we must be able to perform the
following tasks:
Allow any user to create an SMO and a CMO.
Allow any user or set of users to share, store, or modify a CMO.
Ensure that the user to user communications preserves the temporal, logical and spatial
relationships between all CMOs at all users at all times.
Provide an environment to define, manage and monitor the overall activity.
Provide for an environment to monitor, manage and restore all services in the event of system
failures or degradation.
104 | P a g e
We shall see in the next section that the session layer service address all of these requirements.
2.10 CONCLUSIONS AND OBSERVATIONS
We have introduced a new paradigm of viewing information and have developed a world view to
place that construct in. The construct tries to relate the human interface problem and the structure
of both information and knowledge of events into a single schema. We have addressed the issue
of what is multimedia communications, what are the elements of the design process in this new
medium, what are the design dictates of the total environment, and what rules for design can be
abstracted from the change in the experience in interacting with media.
Whatever the ultimate design factors that may evolve, there are certain obvious design
observations that can be made based upon the current understanding of the multimedia area. We
must consider that in designing in a multimedia world we are doing so in a distributed service
environment. It is not a manufacturing business where the product is assembled in a factory
under the control of the owner and where quality can be carefully maintained and managed. It is
a service assembled in the hands of a user in an environment that is neither predictable nor
constant. Thus the design challenge is to incorporate the stochastic nature of the shared user
environment.
To this end there are several design rule that have evolved in the process of understanding such
systems. Specifically:
The next step should always be obvious.
Whenever designing a multimedia system, the presentation to the user, through whatever senses,
should clearly indicate what the next step should be. Where should I type next, where can I place
the pointer, which device do I use, what should I say? All too often designers let the user have
the freedom to create. This results in ambiguity, frustration and visual, aural and tactile
dissonance.
Form matches function.
What are we using the system for and why. What is the function and the form of the system
should match the function that it has been designed for.
There should be consistent paradigms.
When the system is designed to edit images, the editing tools should be the same in all
configurations. The access mechanisms, if they are on the left should always be on the left.
105 | P a g e
Execution should be smooth.
Tactile and visual dissonance are common factors in poor design. A smooth design should be
such as to enhance the conversationality mode of the session.
The question should always be obvious.
State what you want. The statement should be clear and not allow for any secondary
interpretation. If the question is complex, then it should be broken down into smaller segments
and simpler questions. To paraphrase Wittgenstein, the essence of true understanding is the
ability to pose the question in such a way that the answer is clearly yes or no.
The answer should always be obvious.
When answering a question, the answer should always be clear and obvious. Again if a complex
answer is to be presented it too should be segmented.
There should be no ambiguity of expectations.
The users and the designer should have the same set of expectations for the deployment of the
system. "I never thought they would do that with it!" is a common complaint. If all else fails,
listen to the customer, user, etc.
If we can develop systems that follow these guidelines in all of their dimensions, then there
should be a smooth road to transition in the multimedia area.
106 | P a g e
3 MULTIMEDIA ELEMENTS
This chapter discusses the different media elements that encompass the multimedia domain.
Specifically we develop the detailed characterizations of the various media types, developing an
understanding of the issues of how they map into communications and reproduction formats.
One of the issues discussed is the manner in which the media are presently presented and the
transition from what is frequently a mechanical process to and electronic process.
The two issues that we focus on in this chapter are those relating to the characterization of the
overall source in a multimedia environment. Those two issues are the characterization and the
specific sizing of the multimedia image in terms of bits. We know that the bit sizing is essential
for the source characterization and the we shall latter be using it as an integral element of the
system sizing and performance.
3.1 THE MEDIA CONSTRUCT
The main issue discussed in this chapter is how we represent the reality of the observed world in
a form that can be stored permanently and processed at latter times. A simple example is how we
take a picture of a particular person or place. The picture is merely a representation of the person
or place and in many ways does not duplicate all the features of the person or place. We may
then take that film medium and digitize it for storage in a random access memory to be post
processed to either extract features or enhance the image. We can then readily ask the question
of, "How well does the digitized image reflect reality?".
Thus the key element in understanding the use of media for the processing of information relates
first to how we store the information on the different media and second how well the stored
representation relates to the original reality. Media characterization starts with the following
paradigm:
Reality exists and does so independently of the observer. The reality has an extent that can be
projected onto an observer through many means. The representation of the abstract reality is
unique at any one instant and this uniqueness can be represented onto each of the observers
senses in a definable and measurable fashion.
Senses are the major inputs to the human for the purpose of transferring information about
reality. The senses common to the human are sight, sound, touch, smell, and taste.
107 | P a g e
Observations are the results of reality projecting itself onto an observer through their senses.
For example, there may be a representation of a reality in terms of the electromagnetic waves
that is generates and there may be a model for the optical response of the human. The
combination of this quantifiable reality, with this observation system can be further mixed with
We depict this paradigm of the world that we have viewed with an exposition of the example that
we have just discussed. In this chapter we shall assume that we deal only with the senses of sight
and sound and that we shall discuss the different media that can be used to store the information
in a specific form of representation. We further are concerned with the accuracy of that
representation adequately matching the information presented from the reality. It should be noted
that we nowhere deal with relating ourselves to the ultimate reality. This is a very Platonic
approach to understanding our existence but the reader should not read any significant
philosophical positions in this paradigm of a world view.
Medivim Characterization Effectiveness is a measure of the accuracy in which the representation

matches the information. Since both of these are quantifiable we can analytically measure the
effectiveness of the medium in term of the ability to reconstitute the original information bearing
signals.
Thus the two major efforts in this chapter are the ability to understand how information is
reduced to storage on different media and in turn to quantitatively determine the effectiveness of
that medium in retaining a close proximity to the original information.
We define the information of this reality as the electromagnetic wavefront generated by the
reality, projected along the visual senses. Let this information quantity be given by:
(2.1) E(x,y,z,t)
108 | P a g e
Image • The "thing" in itself
Physical Response • The physical "measurement"
Sampling • The conversion to a transmittable entity
Restoration • The restoration to a physical reality
Regeneration • The presentation to a person
Recognition • The human's understanding
We now take a picture of this reality and generate a photoemulsion that can also generate an
observation or an information based image of its own. This information image can also be
represented by an electromagnetic field:
(2.2) El(x,y,z,t)
Now it is possible to measure the error induced between the emulsive image capture medium and
the original electromagnetic information image. We define a measure ||E-El|| as some metric on a
three dimensional space (see McGarty) that allows for the specification of errors. Thus it is
quantitatively possible to measure the effectiveness of medium characterization.
Now we can further take the photographic image an digitize it. To digitize an image we must
perform several tasks. These are typically:
Time sample the wavefront at t(k) seconds, where k+l to N.
For a single sampled wavefront, sample the wavefront in space at point xi and yj. Call the
samples E(i,j,k).
Now quantize the samples of E(i,j,k) into quantized values called S(i,j,k), such that the
quantization is close to the original signal.
109 | P a g e
Now let us assume that we are sampling the picture at the rate of thirty times a second, and that
we have 1 thousand by one thousand element display, that is 1 million pixels, and that we
quantize in one of 1024 levels (e.g. 10 bits) so that we are generating 300 million bits per second
of data on this storage method.
We have now stored the data digitally and we wish to retrieve it. To do so we need a display that
reproduces the image using the stored data files. Let us assume that the display is 512 by 512 in
dimension and can respond to only 8 bits of phosphor response. The image generated and
transmitted towards the human eye is the electromagnetic field E2(x,y,t). We can now describe
how the system can effectively reproduce the image by the error:
(2.3) ||E-E2|| = ||E-G(E1)|| - ||E-G(F(E))||
where G and F represent the transformations of the two media storage methods.
This simple example of using three different media raises the two points that we wish to present
in this chapter. First, we must be able to characterize how primary reality is stored and
transformed by the storage media, and second what the effect of this transformation is on the
overall restoration of reality.
Media characterization starts with a specific reality. That reality, for our purposes, is delimited to
the visual and oral areas. Specifically it is related to what we see and hear. Expanding the reality
to the other three senses (smell, touch, and taste) have not been focused on in the present are of
media communications, however, they surely will evolve as we come to better understand how to
transfer information through these senses.
There are multiple types of information bearing media that are used. Several of the more
common forms are still images, full motion video, and voice. In this section, we develop models
for the characterization of these three types of media and provide a basis for the development of
generalizations for many other types of media.
3.2 IMAGES
The still image is the most common form of display format for Pictorial information. The still
image generally tires to capture some form of reality that is represented, for example by a scene
in nature. That scene in nature may in this simplest form be a view of a tree on a hill with the
clouds in the background.
The first question is how do we take this complex still image and store it in a finite amount of
memory for latter retrieval and display. That problem had been answered over a hundred years
ago with the development of the photograph, a physical/chemical storage medium. The
110 | P a g e
photograph is only an approximation of the true reality however. It uses a form of silver salts that
are responsive to light falling on them and darken to different )degrees. The resolution of such
images is that of the underlying silver nitrate solution and it generally is better than that of the
human eye. Thus the viewer of one hundred year old photographs will not see any degradation of
the image but will perceive a clear and well preserved representation of the scene. The key
observation of this fact is that the eye has a key role in determining an acceptable degree of
image resolution and with such forms of mechanical/chemical memories as 35 mm film, there is
more than adequate resolution.
With the use of computers however and the needs of other media such as the newspapers, there
are lesser degrees of resolution required in image display. This type of resolution is what many
of us see in the print out of our computer terminals or in the daily newspaper. In the case of the
computer printer the image is a composition of small black dots, the density of the dot
proportional to the intensity of the image.
3.3 MEANS OF CHARACTERIZATION
We can now proceed to discuss the details of the characterization of the different forms of
images. We shall take as the starting point, a reality of a fixed scene and from that scene develop
the ways in which we first characterize achromatic (eg black and white) and then chromatic (eg
color) still images.
Before beginning with the detailed discussions of the two major types of images; achromatic and
chromatic, let us first present a brief discussion of the way the human eye responds to images.
The eye is composed of the following elements:
Cornea: This is the outer surface of the eye and generally encompasses the overall outer surface
of the organ. or Iris: This is a circular and pigmented membrane behind the cornea that allows for
the entry of light. It opens and closes in one of its parts, the pupil, upon response to the total
intensity of the light incident on the eye.
The retina is the neurosensorial layer of the eye and its area is about 12J cm2. It transforms the
incoming light into electrical signals that are transmitted to the visual cortex
The anatomy of the retina shows five types of cells organized in layers. The furthest layer from
the incoming light is that of photocells. There are two types of photocells: rods and cones. A
normal eye contains about 130 million rods and 6.5 million cones. Rods and cones are different
enough to be examined separately. Rods are sensitive to shapes and need low luminance
(scotopic vision). In contrast, cones need daylight (photopic vision). They detect color and
distinguish details.
111 | P a g e
Their distribution in the retina is highest in the vicinity of the optical axis of the eye. That is why
a precise detail vision is obtained only when the eye "fix" them, in other words, when their
image is formed at the fovea. In this region there are about 120 cones per degree which fix the
visual resolution to 1 min of arc Photoreceptors are responsible for transforming the in-
The eye is composed of the following elements:
Cornea: This is the outer surface of the eye and generally encompasses the overall outer surface
of the organ.
Iris: This is a circular and pigmented membrane behind the cornea that allows for the entry of
light. It opens and closes in one of its parts, the pupil, upon response to the total intensity of the
light incident on the eye.
Pupil: This is the specific opening at the center of the iris that responds to the light level.
Lens: This is element of the eye and is adjusted by muscle movement to allow for the positioning
of images on the rear part of the eye.
Retina: This is the most sensitive part of the eye. It contains the sets of nerves that allow for the
reception of the light and its transmittal to the brain for processing.
Optic Nerve: This is the connection point between the eye and the never pathway to the brain.
within the retina are two major types of nerves cells; rods and cones. They are called such by
their general nature since cones have a cone shaped head and the rods are generally round with
no tapering toward their heads..
There are 130 million rod type cells in the typical human retina and there are about 6.5 million
cone cells. The cone cells are distributed at about 120 cones per degree of arc which allows for a
resolving capability of 1 min of arc (see Kunt at. al. p 175). Rods are the sensors for the shapes
of objects whereas the cones are the sensors for color and light intensity. The eye responds to
light over the range from 350 nm to 750 nm. The mechanism for this response is the
photochemical responses of the molecules called carotenoids. These molecules are coiled tightly
until impacted upon by photons. Then they rapidly uncoil and when recoiled release electrical
energy that triggers the nerve cells to transmit their messages to the visual cortex. Thus vision is
a complex optico-chemical-electrical process.
The resolving power of the eye is one of the more critical factors that we shall be focusing on in
this chapter. Let us take a closer look at the 1 min of arc number and see how this effects types of
images that we view.
112 | P a g e
Let d equal the distance between two images at the image plane and let A be the angular distance
in degrees of arc between the two points.
d=rA
and we can substitute A of 1 min of arc to this equation. Recall that 1 min equals 1/60 degree,
and one degree is (2 pi/360). Thus we have one minute is equal to 0.0003 radians. At one meter
viewing distance we have a resolution of 0.3 mm. At 0.3 m or about one foot, we have a
resolution of 0.1 mm. We can compare this to a typical television set of 1 foot in height. If there
are 525 lines per scan then there are 1mm between the lines. If we sit 1 foot from the set we can
resolve 0.1 mm or ten times the resolution of the set. However if we view it as 10 feet then we
are at the resolving power of the human eye.
Thus we can see that the issue of better resolution of the human eye to such technologies as
HDTV may be fundamentally specious in that unless we view such system at close range, we
cannot hope to obtain an fundamental improvement.
In a fashion similar to the spatial resolving power of the eye we can examine the fundamental
color perceiving power of the eye also. This has been discussed by Shepherd and others and will
be left to the reference materials.
The eye has six major modalities that impact on how it is effective in multimedia
communications. These modalities are:
Photosensitivity: This is the sensitivity of the eye to the intensity of the light. In particular it is
important in the eyes ability to receive light from diffuse source that reflect light from nondirect
elements. A typical example is the reception of light from movies theater screen which have
marbled surfaces to create diffuse reflections. Without those reflections, we should have a
specular reflection which would be visible to the viewer in the direct reflection direction.
Form Discrimination: This factor relates to the resolving power that we have discussed. The
discriminating capability is dominated by the position of the cone neurons in the retina and thus
limit the resolving capability of the eye.
Movement Discrimination: This is the ability of the eye to detect change. This factor is important
in such areas as film and television. The eye has a time constant on the order on 0.1 second, such
that any events that are shorter than that are not discriminated by the eye.
113 | P a g e
Binocular vision: This factor allows the human to discriminate depth in a spatial context. It is a
function of the two eyes in resolving the differential responses on the retina and the ability of the
visual cortex to process the images and resolve depth.
Polarized Light Discrimination: There is a limited ability of the eye to discriminate in the
polarization of light.
Color Discrimination: The eyes of humanoids are one of few anthropoid eyes that can respond to
color. The color response is viewed as an evolutionary response to the limitation of night vision
in the species.
Thus all of these elements of the eye should be considered when designing and operating a
multimedia system
3.3.1 Achromatic Images
Achromatic images are those that are either black or white, or even mixed with shades of gray.
More specifically, achromatic images do not have any color. An achromatic image is generated
by the response of the eye to electromagnetic waves that are reflected from a surface and
received by the eye. If we have a surface that reflects light at all frequencies equally well, and
that total reflection occurs at the surface, then that surface is perceived as white. On the other
hand if we illuminate a surface with a source of light that has all frequencies and the surface
reflection characteristics are such that it reflects all incident light with zero reflectance then that
element appears black.
Let us consider this black/white occurrence i more detail. First we must remember that we see
objects only as a result of the light that they reflect, or in some cases emit. Let us focus on
reflected light only. Let us assume that the incident light is characterized by a wave of the form
E(x,y,z,t) for all points in space. Further let us assume that the wave can be characterized by a
Fourier transform called E(x,y,z,f), where:
E(x,y,z,t)=∫ E(x, y, z, f) exp(−2 ft) df
If we further calculate the amplitude of the wavefront at each frequency, or equivalently look at
the spectrum of the signal, we see that for white incident light we have:
||E(x,y,z,f)|| = Const
Which is constant for all frequencies or wavelengths.
114 | P a g e
Now let us assume that the reflected wave is given by ER(x,y,z,t) and that we can show that if
the surface has a reflectance at frequency f of, R(f), then;
ER(x,y,z,t) =∫ E(x, y, z, f) R(f) exp(−2 ft)df
Now if we assume that R(f) is independent of frequency, that is:
R(f) = R We then have;
ER(x,y,z,t) = R E(x,y,z,t)
Then, clearly, if the incident wave is white light, we have for:
ER(x,y,z,t) = E(x,y,z,t) (R=l and is white)
or;
ER(x,y,z,t) = 0 (R=0 and is black)
When we discuss chromatic surface we shall see that this changes since R is now frequency
dependent.
When we consider black and white images, we are generally considering that the white light is
reflected from the surface, independent of frequency. This means that the surface has no specific
reflection characteristic. At most, the surface may absorb the light uniformly, thus resulting in a
graying of the light. The grayness in this case is a result of the loss of light intensity. Thus what
we see from a surface is the level of intensity of white light. If we let I be the light intensity, then
using standard electromagnetic principles, we have:
I = | |ER| |2 Using 2.8 we can write;
I - R2 | |E| |
Let us now consider the effects of light on film. Photographic film is simply a gelatin base
interposed with silver halide crystals. These crystals are composed of silver halides of the form
such as AgBr, eg silver bromide. When light shines upon these crystals, the reaction occur:
(2.13) Br> Br + e
)2.14) Ag+ + e > Ag
115 | P a g e
That is, silver is reduced into the gelatin substance. Upon development ( see Jacobson et al) the
developer enhances the result of silver generation in the crystals that have been exposed to the
light. Thus in a region exposed to light, a large set of silver crystals are produced and in those
areas not exposed that are small i number. During development, the remaining silver halides are
washed away, leaving only silver in the developed plate, acting as a darkening agent to
transmitted light. This result in the standard negative that we have seen in many cases. We
should remember that the density of photons is directly proportional to the light intensity so that
we can expand the analysis to what we have discussed in the development Of (2.11).
Now the resolution of film depends upon how well we can generate a suspension of silver halide
crystals, and also how effectively we can develop the film. Such effects as diffusion of the light
and scattering in the film plate may lead to the less than crisp photos we may have often seen. In
the most effective film resolution that we may have experienced, we look at the standard x-ray
film. Here the same film process is use but the incident radiation is at the x-ray frequencies or
wavelengths. The x-ray is used because of its ability to penetrate the hard and soft tissues of the
body. The net result is a film plate that is exposed in proportion to the transmitted, rather than
reflected energy.
The x-ray is a typical example of the continuous gray scale image that is what we commonly
refer to as the black and white photo. Continuous gray scale represents the fines gradation of
exposed silver crystals that can be obtain on photographic file. The resolution of the typical x-ray
film is determined by the density of the silver crystals on the film plate and the ability of the
developer to keep the diffusion of the developer process to a minimum.
If we were to digitize the x-ray film, on the basis of human ability to resolve the resolution of the
image in terms of line resolution and to resolve the image in terms of shades of darkness, a
display of 20" by 20", having 2,000 by 2,000 pixels, or 4 Million pixels per 400 sg in is needed.
This equals a resolution of 10,000 pixels per sq inch. In addition the display per pixel must
resolve 4096 levels of gray, which is determined to be the maximum level of resolution of shades
of darkness of human resolution. Thus a figure of merit for an x-ray is 48 million bits per x-ray,
based upon 4 million pixels, at 12 bits per pixel.
This x-ray example is the simplest example that demonstrates the need for combining the
concept of the original image of the human body, the ability to express it in terms of a digitized
continuous tone image, and the corresponding use of human responsibility as the measure of
comparing the closeness of the digitized image.
This x-ray example has led us into the discussion of gray scale techniques. A gray scale image is
typically a digitized version of a black and white image. Remember that the black and white
116 | P a g e
image is in reality a quasi-continuous display of the intensity either reflected or transmitted by an
achromatic sources, from or through a surface. The gray scale representation is the means of
taking that quasi-continuous image and reducing it to a bounded digital representation. We say
that the original image is quasi-continuous because of the fact that the silver generated on the
film developer plate is microscopic in level and for the most part is as continuous as we can get
in nature. In its ultimate limit all such figures are truly discrete.
Let us continue with our gray scale representation. There are two elements that are of concern in
the digital representation of a gray scale image. They are:
Pixel Resolution: This issue relate to how many pixel per square inch are necessary to be
unresolvable by the human eye. Based upon both psychological and psychological studies, we
find that (see the reference by Kunt et al) the retina can resolve images set apart by 1 min or arc.
Anything smaller than that and the image falls on the same cell. Thus if we have a viewing
distance of about three feet from the screen, we can use the following:
d = r ang
where d is the distance of the pixels on the screen, r is the 3 feet, and ang is the resolutions in
radians. Substituting the parameters we find that d approximately equals 0.01 in or 100 per inch.
This yields 10,000 pixels per square inch. This is exactly what we displayed earlier.
Lightness resolution: In a similar fashion, we have found that the retina can respond to about
4000 shades of light intensity and this yields about 12 bit of shades of gray that can be used.
Thus for gray scale coding, the human eye needs about 10,00 pixels per square inch with 12 bits
per pixel. This means that the typical, continuous gray scale digitization image requires 120,000
bits per square inch at a 3 foot observation distance.
Let us examine the gray scale quantization a bit further. Let I represent the light intensity at a
particular pixel. Let us start with a single bit gray scale quantization. That is at the pixel we use
either black or white. In particular, let 10 be the threshold and say:
P = P0 if I < 10
and
P = P1 if I >= 10
117 | P a g e
Then code this pixel into 0 or 1 according to this level of quantization. The choice of 10 is not
arbitrary, as we shall show shortly. Let us expand this to two bits.
(2.x) P = P0ifI < 10
P = PIif10 <=I <= II
P = P2ifII <=I <= 12
P = P3if12 <=I
We can now code this into two bits. Again the choice of II and 12 are not arbitrary. We can
expand this into 12 bits, or 4096 levels by selecting 10 through 14095. Again the choice is not
arbitrary. We find this through many measurements of human responses and it is also related to
the ability of the retina to respond to stimuli as well as the brains ability to distinguish these
elements.
Now we can view gray scale coding in this form as the ultimate in the achromatic coding or
images from a continuous form. At the other extreme is the mapping of images onto dot matrix
printers that are common in the PC world. A dot matrix printer can produce about 50 dots per
inch or 2500 dots per square inch. Again compare this to the eyes ability to del with 10,000 per
square inch ( see Durrett p. 223). However, the dot matrix printer prints a dot, black, or no color,
white. Thus it is only one bit per pixel. This is simply 2500 bits of information per square inch.
This compares to the eyes ability to deal with 120,000 bits per square inch. This is a 50 to 1
reduction in information.
We can expand this capability with a technique called digital halftoning. Quite simply,
halftoning is a technique that combines the limited capability of having only one bit per pixel
into sets of such pixels that, to the eye, look like more than just one bit, they looks like gray
scale. It is based upon the fact that collections of bits have reflectivity parameters that appear to
the eye as gray scale. The work by Ulichney shows how this gray scale factor can be obtained. In
addition, there are many gray scale algorithms that permit the optimization of the visual
response. These are presented in both Ulichney and the work of Foley and VanDam. Let us use a
simple example to show how halftoning is developed.
We have a set with no dots, one with one, two, three and four. Thus with 2 dots we have 22 or 4
possible entries. Each of these entries have less and less reflectance or in turn have a perceived
shade of grayness. Let us assume that we are taking a gray scale image with 100 dots per inch
and compressing it down to a halftone image with only 50 dots per inch. Let us further assume
118 | P a g e
that we wish to have the halftone image approximate the gray scale image. We shall proceed as
follows:
Use the halftone basis set of the 2x2 matrix and note that since halftone uses only 50 dots per
inch, that two dots will equal 4 dots of the gray scale image. Specifically if we define:
D GS = the gray scale image density in dots per inch
D HT = the halftone image density in dots per inch
N HT = the size of the half tone cell in dots
G HT - the number of gray scale levels in a halftone set
then we can show that we can compress the gray scale into the halftone.
The gray scale compression can be given as follows. The number of gray scale dots compressed
is N GS which equals:
N GS = N HT * (D GS/ D HT)
But in the N GS cells there are L GS gray scale levels per cell which equals L GSB, gray scale
bits. Remember that this is at most 12 bits or 4096 levels. Thus there are a total of:
TB GS = L GSB * (N GS * N GS)
TB GS gray scale bits in the gray scale cells. For example in the case of 4x4 gray scale and 12
bits per cell we have 16x12 or 192 bits per gray scale cell.
However, in the halftone cell, we have 2x2 or 4 cells with only one bit per cell. This yields an
equivalent
TB HT = L HTB * (N HT * N HT)
In this case there are 4 bits. Thus the halftone has reduced the bits from 192 to 4, or a bit
reduction of 48 times. We obviously loose a significant amount in this process, but the resulting
image is often quite acceptable.
119 | P a g e
To select the proper halftone bit pattern, we must look at the several sets of gray scale dots or
cells and average the gray scale level, and then further quantize it to fit the gray scale levels that
have been allowed in the halftone quantization.
This process of halftone production is a well-accepted practice in the field of newspaper print
and in many other printing systems. It is the most obvious choice in use of dot matrix printers.
There are several problems that arise with the halftone imaging however. They typically result
from the use of a set of periodic basis functions to generate the pseudo-gray scale images. This
can be eliminated by the use of entering randomness in the selection of the dot location and in
the assurance of no long run sequences of the same pattern even if the average gray scale density
remains relatively constant.
3.4 CLASSIC COLOR THEORY
Color can be viewed from several perspectives and the two focused on herein are the human eye
and the measure power spectrum. The human eye views color in a complex manner since the eye
receives color stimuli via sets of sensors which are tuned to three possible visible frequencies,
the classic red, green and blue.
We see a flower as a certain color. There is "white" light shown upon the flower and the light is
reflected from the petals and sepals and what we perceive is a red flower. This perception is a
combination of two things; what part of the incident white light is reflected, and how our eyes
process that reflected light. Thus color has two meanings for us. First color is nothing more than
a reflected spectrum of electromagnetic waves in the optical frequency band. Second it is what
we perceive as a human observer and in turn name as a color. The latter approach can and often
is quite subjective. This latter approach is the basis of print color, paints, dyes, pigments and the
like where the end point is the presentation of some artifact of a desired color. It is the former or
first approach we seek to use, namely what is there independent of the observer, specifically the
spectrum.
120 | P a g e
Spectral Characteristics
Power Spectral Density of

Reflected Light
Wavelength
6/5/2008 4
In the context of color as perception there is a collection of terms which should be understood.
Hue is a synonym for a color. Red is thus a hue as is orange. The hues cover the visible
spectrum. Then there is the lightness, ranging from whit to black. The third element is the
chroma which is the departure of the hue from gray. We show these elements below.
Color
Lightness=Black to White
Chroma=Departure from Grey
Hue=“Color”
6/5/2008 8
121 | P a g e
The color school then takes three more steps. These are the CIE models for color. The first step
is the Tristimulus models. The Tristimulus function is shown below. This is NOT a spectrum. In
addition negative values mean more positive stimulus. These are also the result of extensive
experimental modeling. The red, green and blue Tristimulus model as shown below characterizes
three stimuli which affect the three receptors of color in the eye.
RGB Color Matching Tristimulus

Functions
6/5/2008 9
The above Tristimulus curve, called the color matching function, were experimentally
determined through an experiment where a person looked at a test color centered at a wavelength
as shown on the horizontal axis above and tried to match it by adjusting a red, blue and green
lamp in the reference field. This could be accomplished for all regions except between 444 and
526 nm. In that region a red light had to be added to the test field to adjust the color to match. In
effect the test color was changed by adding red. This adding of red is accounted for by the
negative portion of the above curve1.
Following this above model based upon experiment is the spectral approach called the standard
observer consisting of the X Y Z model as shown below. They do effectively represent quasi
spectral responses since they are all positive. It is possible to transform between the RGB and the
XYZ formats.
1
See Berns p. 49.
122 | P a g e
CIE Standard Observer Curves
6/7/2008 12
The CIE Chart is a manifestation of how the three stimuli above can be added together to create a
broad set of colors, hues. It is possible to go from red, thru green and then to blue. One need only
mix the three stimuli in the proper ratio. Then other hues can also be generated.
CIE Chart
6/5/2008 17
To understand this a bit better we analyze the RGB system first. We start with a source specified
by intensity I dependent on wavelength. This allows us to define the following:
123 | P a g e
I ( ) = the spectrum of a specific sample
Define
R= I ( )r ( )d 
G= I ( ) g ( )d 
B= I ( )b( )d 
Note, as we had stated, the color matching primaries show negative values because the negative
was the way the CIE arbitrarily represented an excess positive contribution required to be added
to a primary to achieve the desired spectrum response while keeping the elements normalized.
Specifically:
 r ( )d  =  g ( )d  =  b( )d 
Now this also implies the following are to be true:
R
r=
R+G + B
G
g=
R+G + B
B
b=
R+G + B
and
r + g +b =1
In a similar manner we can do the same for the XYZ system. This is done as follows:
I ( ) = the spectrum of a specific sample
Define
X= I ( ) x( )d 
Y= I ( ) y ( )d 
Z= I ( ) z ( )d 
124 | P a g e
And as was the case for RGB we also have the normalizing factor. Not that it is this normalizing
factor which assures our ability to deal with the triangular plot of color.
X
x=
X +Y + Z
Y
y=
X +Y + Z
Z
z=
X +Y + Z
and
x + y + z =1
Finally there exists a set of transforms which allows one to convert from one to the other. This is
shown below:
r  x 
g  = A  y
   
b   z 
A such that sums of r,g,b and x,y,z are unitary
Therefore for any color C we can write it in one of the following two manners:
C = rR + bB + gG
C = xX + yY + zZ
Thus an x,y plane can be constructed such that any color can be characterized as a pair of
coordinates (x,y). This is the CIE Chart which we have shown above. It must be noted that all of
this analysis is predicated on how "we" see color and not in any context of how it is created or
the underlying physics of color.
Now there are two other brief examples worth noting. First is the concept of additive colors, such
as those we see when we add lights. This was the basis of what Newton did in his early
experiments. By adding lights we can ultimately create white. We show that below.
125 | P a g e
Additive Primary Colors
Additive Colors combine to form white. Traditionally adding lights is

additive whereas adding colorants, pigments or dyes, is subtractive.
6/5/2008 19
The opposite is the subtraction of light, and this is the result of adding pigments of different
colors together in an oil painting. If we were to add all the colors together then we obtain black
and not white. This is subtractive, for we are in reality removing colors by the use of those
pigments. In many ways this is the difference between water colors and oil paint. We show this
subtractive result below:
Subtractive Primary Colors
Subtractive Colors form black. Subtractive mixing involves the removal, subtraction,
of light from the mix. Removing all light ultimately results in black. Absorption only
is called simple subtractive mixing whereas combining this with scattering is
complex subtractive mixing.
6/5/2008 20
126 | P a g e
Notwithstanding the above detail and its use in many industrial processes, these methods used in
classic colorimetry are methods that rely upon the human by necessity being part of the process.
We when looking at plants, shall disregard the human.
3.5 SPECTRA AND MEASUREMENTS
The measurement of absorption spectra can be accomplished by a variety of means. We present

here two methods; classic spectrometry and Fourier Transform Spectrometry (FTS) also called
Fourier Transform Infrared Spectrometry, FTIS. However FTS can be applied to the optical
bands as well.
The goals using these methods are as follows. First to determine the absorbance and extinction
coefficients of the secondary products that are colorants. This means that solutions of purified
anthocyanins, Peonidin for example, would be used and their absorbance and extinction
coefficients determines for all wavelengths over the optical band2. This is accomplished for all
targeted absorbents. Second, perform the same on all known colorants found in a target plant
cell. This could include any secondary product or even proteins which have absorbent properties.
Generally the other chemical elements react in an absorbent manner out of the optical band.
Third, perform the analysis on target cells. Our approach is to perform this on a cell by cell basis
thus requiring focused optical positioning.
3.5.1 Classic Spectrometer
We first look at the classic spectrometer. It uses two paths for transmission, one through a cell
with the target secondary and another cell without any secondaries or colorants. The second cell
is a reference cell. The reason for this approach is to calculate the difference in absorption. The
spectrometer is shown below. It functions as follows:
• Select a Reference and a Sample. Source Spectrum is to be determined using reference.

• Send light from source through both sample and reference. The source must be
broadband wavelength. There will be no need to regulate amplitude across the band since
a difference signal is obtained and the result will be expressed as a ratio.
• Chop the signals using electronic chopper so that half interval it is sample and other half
it is reference. This can be accomplished with a time controlled electronic device or even
a mechanical rotating wheel which can be synchronized to the measurement elements.
• Send to Diffraction Grating to spread out signals over visible spectrum.
2
See Cantor and Schimmel, Part II, pp 380-388.
127 | P a g e
• Sample from one end of spectrum to the other by mechanically sampling the diffraction
grating spread out. Remember that the diffraction grating act as a prism and spreads out
the signal spatially over the optical band.
• Use the reference as the baseline and then measure the ratio or the difference of sample to
reference and plot. This generally requires just a difference amplifier at the measuring
point and synchronizing it with the chopping signal.
•
Spectrometer
Sample
Source
Chopper
Reference
Diffraction
Grating
Detector
6/5/2008 22
The spectrometer, as shown above, functions well for the determination of relative absorption. It
is a long and sometimes cumbersome process because the screen in front of the detector is
scanned slowly and this provides the signal used to ascertain the difference measurement. There
is an issue of accuracy and precision in the collection of data and there is also the issue regarding
the amount of light intensity requires. One should remember that as we spread the light out
through the grating we see the spectrum now spatially but in so doing reduce the signal strength
of each segment. The spectrometer has advantage and disadvantage in this configuration.
3.5.2 Fourier Transform Spectrometer
The FTS is a more recent embodiment of a spectrometer and it eliminates many of the accuracy
and precision issue of the classic spectrometer. In many ways it may be viewed as a mini-CAT
scanner in that it collects data which is the Fourier Transform of the desire waveform, namely
the absorption spectrum.
The FTS works as follows:
128 | P a g e
1. A target sample is placed in front of a detector. The detector is a broadband detector and it
provides at its output the integral of all the power entering across the optical spectrum. The
optical spectrum will be the target spectrum of interest so we delimit the detector to that. We also
assume we know the detector response and that this can be adjusted for by means of signal pre-
emphasis. This means that the detector works as follows:
P =  S ( f )df
Here P is the total power and S(f) is the power spectral density of the combined signal. We will
look at that in some detail in a moment. Now we assume that the detector may itself have a
spectral sensitivity given by H(f). Thus what we really receive if we do not pre-process is:
P =  S ( f ) H ( f )df
Which may bias out answer? The way to avoid this is to do some pre-emphasis on the front end
by using filters which do the following:
P =  H ( f )G ( f ) P( f )df
=  P( f )df
if
H ( f )G ( f ) = 1
This is optical pre-emphasis filtering as one does with FM radio. This is a standard approach.
2. Now let us go back to the input. We assume we have a flat frequency broadband source of
radiation emitted from the source. If now we can also pre-emphasize that as well. Then this
source follows two paths. Path 1 is a fixed path up and down and through the sample. Path 2 is
one that goes to a reflector whose portion is changing uniformly in time and is accurately
measures. This second path then send the same signal with the sole exception that it is phase
offset from the main path. At times it may be totally in phase and at time totally out of phase. For
every position x of the reflecting mirror we measure the combined power spectrum received, the
integral of both signals, measure as their amplitude.
3. It can be shown, we do so below, that if one collects the P values and notes them as P(x) then
if we sample x properly we obtain samples of the Fourier Transform of the S(f) function. Thus
collecting P(x) for the correct values of x and doing so with enough samples we can then
perform an inverse Fourier Transform to readily obtain S(f). This is FTS.
129 | P a g e
Fourier Transform Interferometer
S ( f ) = S1 ( f ) + S2 ( f )
or
S ( f ) = exp(− j 2 d1 f / c) S0 ( f ) + exp(− j 2 (d s + x) f / c) S0 ( f )
Moves from +/- x
Sample
Power Detector Power =  S ( f )df

6/5/2008 24
The details can be displayed simply as follows. The signal received is the direct and the reflected
and they are combined in complex space to account for the phase difference as shown below:
S ( f ) = exp(− j 2 d1 f / c) S0 ( f ) + exp( − j 2 (d s + x) f / c) S0 ( f )
 S0 ( f ) cos(2 fx / c), so that the received power is:
2
P(x)= S0 ( f ) cos( fx)df
c
The result is that P(f) is the real FT of S(f).
FTS shows that we can obtain P(x) and it is the Fourier Transform of S(f) the absorption
spectrum of the sample. We take P(x) for many values of x and then inverse FT.
2.2.1.2 Chromatic Images
The ability of the eye to discriminate colors is one of the factors that discriminate the human
from the other animal species. The achromatic discrimination abilities relies upon the ability of
the eye to be sensitive to the intensity of the radiation of the surface of the retina. The color
discrimination capability is the ability of the same neurons to respond in a differential fashion to
photons of differing wavelength.
130 | P a g e
In this section we shall develop the concept of chromatic images, and show how those image
are represented and formed and how we can characterize those images for the ultimate purpose
of sizing the source and storing the data. The normal visible range of the eye is from 450 nm to
650 nm. Some humans may respond to slightly lower and higher wavelengths. The typical sets of
standard colors are :
Pure Blue 470 nm o Pure Green 505 nm o Pure Yellow 575 nm o Red > 610 nm
The concept of pure green is really a psychometric factor. It states that when this color, defined
in terms of a narrow linewidth source of radiation, is shown to a large group of people, then they
agree in large numbers that the wavelength shown is the pure representation of yellow to them.
In contrast, there does not seem to be a uniquely agreeable wavelength for red.
We can define the colors that we have seen in terms of the source of the light if the light is
transmitted directly into the eye or of the reflective nature of the surface for light that is reflected
into the eye. Let us consider first the issue of transmitted light and then reflective light.
If we use a narrow laser source, we can define the transmitted light in terms of the power density
an x,t function of wavelength. Define P(l) as this spectrum and we have:
P( ) = Po f( )
where Po is the peak intensity and f( ) is the spectral characteristic. Light is said to be saturated
depending upon how broad a bandwidth the source is. Thus a narrow band light source is said to
be more saturated.
We can in a similar fashion look at reflective surfaces and consider the effects of white light
being reflected from the surface and entering the eye. The material on the surface will have
differing reflective characteristics depending on the wavelength. Let Po be the intensity of the
incident radiation, and we shall assume that it is white light. Let R(L,x,y,z) be the reflectivity of
the surface as a function of the wavelength and the position. Then we can define the light
incident on the eye in the standard fashion. These types are subtractive and additive.
The subtractive techniques takes white light as the source in a transmissive fashion. It then
passes the light though filters that take out narrow bands of light and leave a combination of all
other colors. Here we have shown white light on a yellow filter that passes right that has had
yellow removed and then we pass it through a filter that has a capability to remove cyan. The
resulting color is green as perceived by the observer. It should be noted that this perception
131 | P a g e
depends highly on the narrowness of the filters and the ability of the filter to be centered on the
desired wavelength.
Consider now the additive form of color mixing. This is the more common form and the one that
we shall be using at length. In this case we take three primary colors, red, green and blue and
start with equal intensities of each. We then vary the intensity of the separate colors to create the
total color. The resulting total additive color is:
IT = I(R)+I(B)+I(G)
The resulting colors can be a wide mixture of all colors perceived by the human eye. Here we
have shown somewhat broadened spectra. We then combine two of the spectra and generate a
third combined spectrum that peaks somewhere in between the two. It is this effect that allows
for peaking at any point and thus having the psychometric effect of a continuum of colors.
The recognition of this additive effect of emitted spectra and the generation of continuous color
has led to various characterization of color mixtures. The simplest is the chromaticity chart that
starts with the additive equation of the sources and normalizes the overall intensity. That is we
start with the equation:
T = 1 + R+B+G
so that we can define :
B = 1-R-G
Now we can plot the amount of red and green on a two dimensional chart, knowing that the
amount of blue is merely subtractive.
over as large a range as possible. Calculation of the luminance of a color would be made much
easier if the luminosity coefficients of two of the primaries were equal to zero. The luminance of
a color would then be equal to the number of units of the other primary used in the match.
It was these considerations, among others, that led the CIE to propose a new set of primaries.
The spectral locus is totally contained within the triangle formed by these new primaries denoted
by (X), (Y) and (Z) (as can be seen in Fig. 5) implying that all spectral colors can be matched
with a positive quantity of each primary. The use of such "nonphysical" primaries should be no
cause for concern. For measurement purposes any real set can be used and the results can be
transformed by a 3 X 3 matrix to the nonphysical set.
132 | P a g e
Luminosity information is obtained from the tristimulus value of the (Y) primary-the luminosity
coefficients of the other two primaries are equal to zero. The resulting tristimulus values 7, 7 and
T are shown in Fig. 7. Note the all-positive nature of the functions, and that y is identical to the
Vx curve of Fig. 2. The xy chromaticity diagram for the 1931 CIE Standard Observer is shown
in Fig. 8. The equal-energy white (£) has the coordinates (1/3,1/3) because it is the reference
white for this system. All the color mixture properties that we have previously described for the
rg diagram are valid for this diagram, however, the equations for color mixture are especially
simple for this case. If we are given the chromaticity's and their luminance's Lx and L% the
chromaticity of the mixture is simply that is illuminated by a light of a specific spectral
distribution. The spectral distribution of the reflected light may, of course, be thought of as being
composed of an infinite series of spectral color. To determine how much of each primary is
needed in the mixture, a product is formed of each of the tristimulus values with the spectral
reflectance of the object as shown in Fig. 6. The areas under each curve, as obtained by
integration, are the desired tristimulus values R, G and B for the sample.
In 1931 such calculations were commonly performed on desk calculators, and the negative lobes
of the functions of Fig. 4 introduce negative product terms in which the negative sign is error
prone with repetitive summing and differencing operations. It would be much better if there were
no negative lobes, and it would be convenient if the quantities were zero
We can then generate any desired color by combining the amount of red or green on this
normalized basis. Note that at R=G=0 we have pure blue. In addition we can note the extension
of this chart beyond the region for the sets of colors generating them. This is a measure of the
psychometric factors generating these colors.
The implications for the use of this psychometric definition of color and the human response
falls into many categories. Two of them are most important in the areas of multimedia
communications. The first is the use of color in the RGB combinations for the generation of
multiple colors on a CRT screen. The basic CRT screen is in actuality a set of overlays of Red,
Green and Blue phosphors. Using these combinations we can generate any set of colors on a
CRT screen. Consider the types of CRT displays (Merrifiedl p.70). The RGB phosphors are
located in a spaced form on the surface of the screen. Three different electron guns are displayed
behind the screen and are focused through a set of masks. These masks are called shadow masks
and ensure that the guns of the different colors do not impinge on the wrong phosphor.
The resulting image is a collection of the RGB phosphors that are illuminated by an intensity in
proportion to the color that is desired at the specific location. The granularity of the shadow
mask is a measure of the pixels per inch of the display. For example, the distance between the
Red phosphors on the horizontal or vertical axis may vary from 0.6 to 0.2 mm. The discrete pixel
y* =
133 | P a g e
is the combination of a set of pixels. If we connect an RGB combination in a triangular fashion,
then we can see that with 0.3 mm spacing, we see that the pixel of color, an RGB triangular
combination is on the order of 0.2 mm. As we discussed in the resolving power of the eye, this
gives significant resolution at close distances.
Specifically, with 2,000 by 2,000 monitors, we can have 6,000 by 6,000 individual elements of
color at the RGB level yielding the overall combination of 4 million pixels. With a 60 cm by 60
cm display, we have 3000 RGB cells at 0.2 mm spacing. In the current state of the art we have
0.1 mm spacing of these cells.
The shadow-mask color CRT assembly consists of three closely spaced electron guns, a shadow-
mask, and a three color phosphor screen. Focused electron beans emitted from each primary gun
pass through apertures in the metal shadow-mask and impinge upon phosphor dots for each
corresponding color. The three electron guns are arranged in an equilateral triangle, or delta. Each
shadow-mask aperture allows the three electron gun beams to project onto an inverted delta or
triad of phosphor dots. The angle of incidence of an electron beam as it passes through a shadow-
mask aperture determines the color of phosphor dot it excites. Electron beams of a particular gun
are blocked by the shadowing of the mask from impinging upon the other two color phosphor dots
of each triad. A shadow-mask CRT has a very simple mechanism for selecting color. The three
independent guns in the shadow-mask design provide independent control of the luminance of the
red, green, and blue phosphors. In this manner, it is possible to reproduce any color within the
chromaticity triangle formed by the primary colors.
Several configurations of gun alignments, mask structures, and phosphor arrangements are
available. The in-line gun shadow-mask
CRT has the mask and pi the in-line gi through a sh cause of the Resolution c to their small beam
at the also available vertical pho a higher lurr The granular distance best ranging from are
considered to be confused several mass electron optics.
The ways of storing still images has evolved significantly over the past decade. The method most
used ten years ago was that of paper tape and today there is extensive use of laser disk storage
technology. In this section, we will develop some of the current and some of the potentially
significant means of storing and retreading still images.
As part of the storage process, we first want to consider the means of compressing the image data
and as part of that understanding the ultimate information in the image itself. This latter issue
relates to the minimum number of bits that are necessary for representation and storage of the
image.
134 | P a g e
The issue of reconstituting still images is one of obtaining either the best video display
techniques or in using many of the existing and improving methods of hard copy display. We
often think of reinstating the image on a video display and do not recognize that the use of hard
copy display terminals is also a significant means of reconstituting the image. There are dramatic
advances being made in the areas of both displays and hard copy devices and these will be
reflected in the way the systems will evolve with the capability to provide significant editing and
composition interactivity on displays.
As we have discussed with the set of still images, the key factor in developing a means to display
images is finding an acceptable tradeoff between the ability of the human eyes to respond and the
limited processing and storage capability of the computer. In the world of video images, there
again is the resolving capability of the human eye, combined with the issue of the field of view
of the image and the update rate of the images to meet the eyes ability to comprehend motion.
Thus the selection of the size of the standard television set in terms of its aspect ratio was based
upon significant testing of the viewers ability to view information in that form. The update rate
of Television is 30 frames a second, a rate that matches the lower limit of the human response
curve for recognizing image motion.
3.5.3 Video Means of Characterization
The video image is more than just the sequencing of a set of still images. There are several
psychological and physiological factors that relate to the scan rates and interlacing of images to
provide an adequate video image. In this section, we will focus on the scanning input to the
video display and shall also discuss the scanned output of the display.
The video displays follow techniques that extend those of the CRT displays that we developed in
Section 2.2. Specifically, we scan the screen in frames, in the US and other countries using
NTSC video, this is 525 lines of scans per frame and 30 frames per
second. A frame is composed of two fields, the first field being 262.5 lines and the second field
being a same number of lines but these lines spaced between the first field lines. The two fields
compose a single frame.
The signal that is generated at the transmitter is a combination of the red, blue and green values
of the image that is scanned at the original location. This is not the signal that is transmitted.
What is transmitted is a set of three video signals that is a linear transformation of the RGB
signal set. This composite transformation set is generated so as to make the received signal
backward compatible with the standard black and white sets that were in existence in the United
States at the time of the introduction of color TV. In addition the signals are required to fit into
the same bandwidth as the black and white signals.
135 | P a g e
If we let R, G, and B be the red green and blue signals, then in NTSC video we generate a signal
set that is composed of the following three signals;
Y=0.3R+0.59G+0.11B
I=0.6R-0.28G-0.32B
Q=0.21R-0.52G+0.31B
These signals are called the brightness or luminance signal, the inphase signal and the quadrature
signal. It is found that with this transformation and the scan rates selected for capture and
ensuing reproduction of the signal, that Y occupies 4.2 MHz of bandwidth and if processed by
itself provides an adequate black and white signal. I occupies 1.5 MHz of bandwidth and carries
the orange/cyan mix of colors in its signal. The Q signal is 0.6 MHz in bandwidth and carries the
green/purple colors of the signal.
It is actually possible to display these signals by themselves, that is the I or Q signals and see
what the resulting image appears like.
The composite video signal of Y,I and Q are carried in a complex fashion. The Y signal occupies
the lower end of the 6 MHz bandwidth spectrum and is not encumbered by any other signal. The
I and Q signals are carried at a frequency that is 3.58MHz up from the carrier. The I signal is sent
using a sine modulation and the Q signal using a cosine modulation. Coherent demodulation is
used at the receiver so that the two signals may be separated. At the very upper limit of the band,
or the carrier plus 4.5 MHz is located a small FM voice carrier that is used to include the voice
signal. All of the video signals are carried in an AM format.
At the receiver, the signals are recovered and the Y, I and Q signals are placed through a
transformation to recover the R,G and B signals again. These are then used to excite the three
phosphors on the television screen.
Storage of video information is a significantly greater challenge than that of the still image. The
still image storage is for a single or set of single images, whereas the storage requirements for
video are based upon a continuum of many such frames at the scan rate of the device. In this
section, we shall first characterize and size the storage requirement and then provide a set of
alternative storage systems.
The means of reconstituting a video image is usually performed on a video display. In this
section we hall concentrate on the display architecture and show how we can implement many of
the schemes that we have discussed.
136 | P a g e
The human voice is a complex information transmission medium that is created with the
interworkings of the human brain, overall nervous system and the vocal chords of the human
body. Combined with the creation of speech is its reception by the human ear which has its own
characteristics.
3.5.4 Voice Means of Characterization
The voice is the second of the sense that we find in a multimedia system and its characterization
can take on several dimensions. First, we can look at the signal itself, that is the conversion of
the sound pressure wave into and electrical wave, and develop the digitization of that signal.
From that digitization we can determine the bandwidth necessary for adequate transmission and
size the storage requirements. A second approach is to recognize the speech using many of the
speech recognition systems and then convert the speech into a new form, the written word. With
the developments in speech recognition it is now possible to do this type of characterization for
many applications.
Let us first develop an understanding of the generation and reception of speech in the human.
The basis elements of the speech generation path are:
Oral cavity including the tongue and mouth, within the confines of the lips and down to the
velum at the rear of the mouth.
Nasal Cavity from the nostrils at the front to the pharynx at the ear.
Oral pharynx from the velum down to the epiglottis,
Nasal Pharynx at the rear of the nasal cavity and above the oral pharynx.
Laryngeal pharynx region from the velum down to the vocal chords.
Each of these areas are resonant regions that are used for the generation of particular types of
sounds. The vocal chords are used in particular in the generation of speech by passing the air
from the lungs through the chords and by exerting various modulations o the chord openings by
the local chord muscles. This air can then be made to resonate at different modes in the various
cavities and the resulting sounds are speech. If we were to block any one of the resonant cavities,
all other elements working properly, we would seriously affect the sound. This is a common
effect when a person has a cold.
The roof of the mouth can be divided into two principal regions. In front, the roof is formed by a
bone called the palate which separates the mouth from the nasal cavities and supports the upper
137 | P a g e
teeth. At the back of the palate, the roof is formed of muscle and connective tissue; this structure
is called the velum, or soft palate. (Writers who call the velum the soft palate refer to the palate
as the hard palate to distinguish the two.) The uvula is a small fleshy appendage at the rear of the
velum. The velum can be lifted by a muscle and pressed against the back wall of the pharynx to
seal the nasal passages off from the rest of the vocal tract. At the front of the palate there is a
ridge, formed by the thickening of the bone where the front teeth are inserted; this is called the
alveolar ridge.
The tongue is a large system of muscles connected in front to the lower jaw and in back to bones
in the throat and head. Its biological functions include tasting and manipulation of food during
mastication. For our purpose it is convenient to divide the tongue into regions. In the absence of
distinct landmarks, it is difficult to define these regions precisely, but they are approximately
The ear has three major parts; the outer ear, the middle ear and the inner ear.
The outer ear extends from the pinna or outer ear surface to the eardrum through the meatus.
This is itself a resonant cavity for the propagation of the sound entering the human ear. The
middle ear is compose of the ear drum, the bone structure called the ossicles and the entry into
the Eustachian tube. The ossicles contains three bones called the hammer, anvil and stirrup.
These bones are used for the transmission and conduction of the sound responses into the nerve
endings that are key to the transduction of sound into the overall nervous system.
The inner ear contains the cochlea which converts the mechanical vibrations into electrical
impulses in the hearing part of the central nervous system.
The human ear can hear sounds from the range of 16 Hz to 16 KHz with some exceptions up to
20KHz. The sound pressure limits range from 0 dB or 0.0002 dynes/cm2 to 130 dB. At the high
level the human ear generates pain and is damaged. At the lower level there is no perceptible
response.
3.6 HEARING AND PERCEPTION
By hearing we mean the process by which sound is received and converted into nerve impulses; by
perception we mean, approximately, the postprocessing within the brain by which the sounds
heard are interpreted and given meaning.
3.6.1 Hearing
We will start with the anatomy of the ear. The ear is divided into three parts: the outer ear, the
middle ear, and the inner ear. The outer ear consists of the pinna (the visible, convoluted
cartilage), the external canal (external auditory meatus), and the eardrum (tympanic membrane).
138 | P a g e
The pinna protects the opening; its convoluted shape is thought to provide some directional cues
(Schroeder, 1975). The external auditory meatus is a nearly uniform tube approximately 2.7 cm
long by 0.7 cm across through which the sound passes to reach the eardrum. Like all tubes, it has a
number of resonant frequencies, of which only one, at approximately 3 kHz, falls in the frequency
range of speech. The tympanic membrane is a stiff, conical structure at the end of the meatus. It
vibrates in response to the sound and is the first link in a chain of structures which transmit the
sound to the neural transducers in the inner ear.
The middle ear is an air-filled cavity separated from the outer ear by the tympanic membrane and
connected to the inner ear by two apertures called the oval and round windows. The middle ear is
also connected to the outside world by way of the eustachian tube, which permits equalization of
air pressure between the middle ear and the surrounding atmosphere.
The middle ear contains three tiny bones or ossicles which provide the acoustical coupling
between the tympanic membrane and the oval window. These bones are called the malleus
(hammer), incus (anvil), and stapes (stirrup). The malleus is attached to the tympanic membrane,
the stapes to the oval window, and the incus connects the two. The function of the ossicles is
twofold: (1) impedance transformation and (2) amplitude limiting.
The signal generated by the human voice system generates a pressure wave of the form:
p(x,y,z,t)
which is spatially and temporally dependent. The wave is received by the ear and is processed
into recognizable speech. We frequently can take this pressure wave and receive it by an
electrical transducer like a microphone and convert it into an electrical signal. The resulting
signal is a representation of the temporal characteristics of the speech. It does not provide full
spatial characteristics and in fact involves the impact of the directionality of the receiving
transducer.
As we have already noted, speech is constrained in the 16 Hz to 16Khz range so that the speech
signal:
s(t)
is composed of only those specific frequencies.
139 | P a g e
If we define s(t) as the speech signal, we say that this signal is a random process since it changes
in time in a random fashion. We can determine the correlation function of this process by talking
its average as it is mixed against a shifted version of the signal. We define R(T) as the correlation
function of the signal. Specifically if we let E[ ] be the averaging operator on the random
process, then we have:
R(T) = E[s(t) s(t+T)]
where we have assumed that the process is stationary, that is R is independent of t.
The spectrum of the signal is defined as the following:
S(f) =∫ R(T) exp(−j2 fT) dT
The spectrum for a typical voice signal is shown below
140 | P a g e
We can now use the Nyquist sampling theorem that states that if we sample a signal at twice the
maximum frequency rate, then we can totally reconstruct the signal. Thus if a signal is from 0 to
l0 KHz, we must sample at 20,000 per second. For voice this means for full fidelity we must
sample at 32,000 per second. Generally this is not the case and we frequently limit the upper
voice signal in telephony rates to 4 KHz and the sample rate is 8,000 per second.
We can now quantize the samples to create a digital sample. If we quantize s into two levels all
we need is on bit (eg 0 or 1). If we quantize into four levels we can do that with two bits. If we
continue, we can quantize into 256 levels and do this with 8 bits. This 8 bit quantization has been
shown to give minimum human distortion in terms of the human response.
Thus typical speech is sampled at 8,000 times per second and at 8 bits per sample. This means
that voice is captured at the rate of 64,000 bits per second. High fidelity speech may be at the
32,000 sample rate and may even use 10 bit quantization. Thus for high fidelity speech we have
320,000 bits per second. Let us review this in further detail.
First we have the general signal s(t) which is the voice signal.
Second, we take and sample the signal at the sample points. The sample pints are twice the
maximum frequency of the spectral density. The corresponding sample time intervals are defined
as T sec. For 4 KHz bandwidth this is 1/8,000 sec or 125 micro second sampling.
Third, enumerate the sequence of samples;
141 | P a g e
(s(kT) } : k = 1, , N
as the sampled signal.
Fourth, using a quantization scheme, select the number of quantization levels, using an a priori
choice of through an optimizing process, but choose the level to be consisted with a binary
format. That is the quantizing levels should be some factor of 2 to a power. Let us assume that
there are 256 levels or 8 bits of quantization.
Fifth, choose the quantizing levels either a priori or through an optimizing scheme. The levels
are defined as follows. Let Lk be the kth level and let the total level be;
{Lk} : k=l,...,2N
where N is the quantization bits.
Then we have the quantization mapping of :
s(kT) is in Ln if Ln< s(kT) < Ln+i
Sixth, then map s(kT) into bit pattern Bn which is the binary level of Ln.
142 | P a g e
Seventh, create a data stream of the form:
[B1,B2,,Bn,Bn+l,]
where Bi is the specific bit stream:
Bi = [0,1,0,1,1,...,0]
3.6.2 Voice Sampling.
Voice storage systems are generally well developed and the major factors in such systems are the
ease of access and the ability to store large volumes of speech while still retaining the quality of
the original.
3.6.3 Means of Reconstitution
Speech regeneration can be done in one of two ways. As we discussed earlier in the section, we
can take the digitized version of the speech and regenerate it directly by reversing the process. A
second approach of speech generation from the written word, this is called speech synthesis. The
development of speech synthesis systems has been significant over the past several years and
these now provide interesting alternatives for use in multimedia systems.
143 | P a g e
144 | P a g e
3.7 GRAPHICS
Graphics is distinct from the other three representations. In still image, video and voice the
information medium is naturally created and is generated from some form of human activity. In
145 | P a g e
the graphics world, the information is created by a computer to represent some form of reality
but is not a priori created by the natural world. Graphics development has progressed
significantly over the past few years. The original systems were those that were developed to use
a minimum of storage and a minimum of processing. In addition they were designed for vector
oriented displays and not for fully bit mapped displays. In this section we shall discuss three of
the graphics standards and show how they have evolved through what is now their present state.
The first standard that we shall discuss is the GKS, Graphics Kernel System, standard that was
originated in the 1970s. It is the oldest standard and many of its characteristics reflect the
limitations on the technology at the time it was developed. The standard uses the basic vector
graphics approach and although it is designed for use on many machines, it does not press the
capabilities of the processors to allow for maximum flexibility. Thus it has only two dimensional
characteristics and requires significant detailed definition for a fully flexible design.
The second standard is the PHIGS, Programmer Hierarchical Interactive Graphics Standard, that
is much more flexible and assumes an generally more capable processor. It however still relies
on the vector graphics approach to display.
The third standard, although a defector standard, is that of PostScript. This system was
developed in conjunction with the work on the Apple Mackintosh system and takes full
advantage of the bit mapped displays. It functions on smaller machines, utilizing the power that
is available in the smaller processors, smaller in size but significantly greater in processing
capability. In addition is recognizes the availability of the bit mapped display and the inherent
advantages of the existing MAC display software and the introduction of laser printers.
3.7.1 Means of Characterization
Graphics characterization has evolved over the past ten years with significant growth in
capability. The graphics software packages are now extensively used in engineering,
architecture, and in many artistic applications to design both still and full motion depictions of
real and imagined events. Unlike stills, video and voice, the relationship to the external reality
for this media characterization is less strong. For example, we can think of a designer who is
developing a new auto body and in so doing is using a graphics design package to help do so.
The graphics package starts with the ideas that the designer has and follows the designers lead in
developing the body design of the auto. In this case, the output is the reality and not the input.
Many standards have been developed for the development of graphics displays. These standards
rang from such as GKS, the Graphics Kernel System, to PHIGS ( the Programmer's Hierarchical
Interactive Graphics Standard), the CGM (Computer Graphics Metafile).
146 | P a g e
Graphics systems take a set of from desired images and reduce them to a defined set of primitive
elements that in sum make up the resulting image. Let us consider a simple example. We may
desire to develop the graphical representation of a computer terminal. We can first start with a
set of rectangles that can represent the overall structure of the terminal, the key board and the
display device. We can then add to these basis elements sets of rectangles, circles or ellipses that
make up the keys, displays, logos etc that make up the final terminal. We can then apply further
and further detail to the device to make up a more complex and detailed display. We may even
use the capability of creating complex curved surface by using spline functions that are basically
higher order polynomials defined on a limited set of R N.
The net result of a graphics system is a picture representation that can fall into one of two major
categories, vector graphics or bit mapped. In the bit mapped systems of generation, the images
are created by a fully random selection of bits and color associated with the pixels. That is we
can imagine that the end user may be capable of entering in to the system the image on a bit by
bit basis and doing so by selecting a pixel and then choosing the colors or intensities for the
pixel. This would mean that for a fully bit mapped system, we would have to have N by N
entries with the B bits chosen per entry. Thus there would be N*N*B steps involved. For the 2K
monitors, this would entail almost 100 million steps per screen. Such an endeavor is beyond the
capabilities of most humans.
The alternative is to select a system that has a set of basis functions such as circles, rectangles,
ellipses, etc and that in addition is aided by a set of primitive commands that allows for the
placing, orientation and coloring of the basis functions. The net result is a vector oriented system
that does not require a bit by bit painting of the display device. In addition this system of basis
functions and primitives allow more readily for the animation of the graphics images.
Graphics standards have several objectives in the development of their design. Typical sets of
objectives that will be found in the three systems that we are to discuss are as follows:
1. Machine Portability
2. Application Program Independence
3. Display Independence
4. Hierarchical Designs
5. Ease of Compilation
147 | P a g e
We shall see that some of these goals are not fully achieved in all of the approaches but that they
are generally thought of in the implementations.
3.7.2 GKS
GKS, the Graphics Kernel System, is a graphics design tool that has the status of being the first
graphics standard. It is composed of six general elements:
1. Primitives
2. Attributes
3. Attribute Bundles
4. Segments
5. Viewing
6. Input
Let us describe each of these in turn.
Primitives: The primitives or output primitives, or output functions are those functions that are
defined in GKS that are used for the generic characterization of the basic sets of elements that
are to be drawn from in GKS. The primitives consist of the following set:
Polyline: This allows for the drawing of a set of connected lines from any point to any other set
of points.
Polymarker: These are markers that are placed on the figures to specify particular locations on
two dimensional surfaces.
Text: This allows for the display of arbitrary strings of characters.
Fill Area: This allows for the defining and filling of arbitrary boundaries of polygonal figures.
Cell Array: This displays a grid of rectangular elements.
Generalized Drawing Primitive: This primitive allows for the generation of arbitrary shapes of
figures.
148 | P a g e
Attributes: Attributes are specific characteristics than can be attributed to any of the primitives.
The attributes are specified in the form:
SET FILL AREA COLOR INDEX
SET POLYMARKER COLORINDEX
SET TEXT ALLIGNMENT
For example the attribute SET LINEWIDH SCALE FACTOR affects the polyline line width
and allows for arbitrary scaling.
Attribute Bundles: These are collections of attributes that allow for the defining of specific
primitive attributes out of the main source code. For example we can use the expression:
SET POLYLINE INDEX
to set the polyline to the characteristics of index 1. Index one may be machine dependent and the
applications programmer may not have to worry about the portability of the code on a machine
dependent basis. Thus GKS allows for the assignment of attribute and primitive bundling
through the index function.
Segments: Collections of primitives and attributes, and even attribute bundles create segments,
which are separate descriptive parts of the total graphics picture. For example if we are drawing
a house, we can create a segment that is the roof of the house, composed of many polylines, and
their attributes and many of the other primitives as necessary. The net result is the segment of the
graphics representation.
Viewing or Transformation Functions: This element of GKS allows for the definition and
description of the viewing space of the GKS display. Typical viewing functions are:
SET VIEWPORT SET WINDOW
SET WORKSTATION WINDOW
These functions are used to specify such functions as the location of the window coordinates on
the screen.
Input: This function provides for the operator input to the program and allows for the entry of
such elements that can provide location, select numbers and text for input, and select or pick
149 | P a g e
specific segments of graphics for display or inclusion. The input functions are in effect a
command language for the operation of the GKS kernel.
It has three main layers. They are:
1. Primitive Attributes that are used to characterize the primitives and create the individual
segments of the GKS image.
2. Segment Attributes that are used for the detail specification of the segments.
3. Workstation Attributes that provide the details associated with the specific display
device. These are typically given in terms of specific device drivers.
A typical GKS image generation shows how the sets of primitives scan be combined to generate
a set of segments. These segment in turn are used to generate the overall image. GKS
descriptions are generally machine dependent and are structured by the collection of segments of
the image. GKS allows for device independence by allocating the device specific portions the
device drivers that are separate from the source code.
We can now take any GKS image and address the issue as to how we characterize the image in
terms of data storage. Specifically, let us consider the image of a house. The hose is composed of
the following segments:
1. Roof
2. Chimney
3. Body of House
4. 10 Windows
5. 10 Window Shutters
6. Door
7. Light
The attributes associate with the roof are those of color, texture, scale and several other factors,
similar attributes are attributable to the other primitive in the segments.
150 | P a g e
Now we can look at the image from two points of view. First is we look at the 1,000 by 1,000
monitor at 12 bits per pixel, we have
12 million bits for the display. If on the other hand we look at the segment characterization, we
can say the following:
1. 7 Separate Segments
2. 10 primitives per segment
3. 20 Attributes per segment
4. 20 Characters per segment
5. 8 bits per character.
This yields 244K bits for the storage of this image using GKS and no coding for either primitive
or attribute expressions. This is a 50:1 image compression based just upon the fact that we know
how the image is constructed. This is typical of many of the image systems that are found in the
graphics area.
3.7.3 PHIGS
PHIGS is a second generation graphics standard. PHIGS is dramatically different from GKS in
many dimensions, the two most important being the hierarchical nature of PHIGS and the second
being the ability to dynamically center attributes as the application program is being run.
Here we first show the basic application program which the PHIGS terms are imbedded. This
interfaces with the PHIGS package that includes the graphics system to actually generate the
graphic images and its associated graphics data elements.. These are then interfaced with the
Input device drivers and the graphics display drivers that finally interface with the operator. As
with the GKS approach, there is a segmentation of the device dependent elements and the
imbedded image dependent elements.
PHIGS has the two basic elements of the primitive and the attribute. The primitive carries with it
the same function of form and shape that we had in GKS. Thus the primitive of a line or circle
can be found in the PHIGS format. The attribute is similar to that of GKS in that is associates
with a primitive a specific set of characteristics such as shading and width.
The difference between PHIGS and GKS is that in GKS we bind the attribute and the primitive
together once and for all. In PHIGS, we can associate a set of attributes and primitives together
151 | P a g e
as needed and this association can change with time. The binding of the two elements is an
output process not a definition process.
This allows for a significantly more flexible design and it allows for the ability to perform
animation in a three dimensional context which is not possible with the GKS type of design. Like
GKS however it is a structural language and builds from the vector graphics paradigm.
PHIGS is a hierarchical structure. Specifically it allows for the development of

STRUCTURES that are bindable primitives and attributes and these structures can be composed
upward or decomposed downward. For example in GKS the HOUSE was defined in terms of the
roof, the chimney the windows, etc. In PHIGS, we can take the house and take any one elements
and further decompose it. Thus the roof can be decomposed into the shingles, the nails, the edges
etc. The shingles can be further decomposed into the tar, the paper, and the gravel. We could
continue this process
down to whatever detail is necessary. This approach in PHIGS allows the graphics designer to
provides greater detail in the specific image.
3.7.4 PostScript
PostScript was the development of John Warnock while at Xerox PARC and is focused on the
truly bitmapped display devices such as high resolution terminals displays and laser printers. The
PostScript approach is one that mimics the paradigm of the printer placing ink on the page to
obtain the effect that is sought by the graphic artist. It is a much more flexible development
language and implementation system than many of the other alternatives. It, in addition, is
directed at the transfer of the developed graphics to paper, whereas the PHIGS approach is
optimized for the continuing display on electronic media.
The basis design object in PostScript are the:
1
o Text Elements that provide the large selection of standard and customizable text fonts.
Geometric elements that provide for the definition of circles, lines and rectangles.
Sampled Image elements that permit the direct importation ad manipulation of external images.
These are all combined in the overall construct of the imaging model. This is the equivalent to
the segment of GKS or the structure of PHIGS. The imaging model is composed of three
elements;
152 | P a g e
Current Page: This is the abstraction of the working space on which the graphics layout is made.
The current page is the blank space that is filled out by the designer.
Current Path: This is an abstraction that may be independent of the current page and it is the
collection of the basic graphics objects.
3.8 CONCLUSIONS
This chapter provides the reader with a method for the characterization and modeling of various
multimedia information sources. The source characterization problem is one of the key problems
in the development and analysis of multimedia systems. The source of a single user must be
composed of a set of the fundamental sources that that user may interact with. Those
fundamental sources are those that we have developed in this section, and possible others that
may be developed as new technologies are developed.
The most significant fact that the reader should obtain from this chapter is the recognition that in
a multimedia environment, we are always striving to duplicate or record a reality that exists
external to the machine. The techniques for doing so have been developed both within the
context of the computer world and within the world of print medium. We too often forget the
importance of the print medium as a means of transferring information from one location to
another and from one person to another. We also are seeing that with the introduction of such
applications as digital printing systems, we are seeing the melding together of the print and
electronic industries from both the operations and applications viewpoint.
153 | P a g e
4 MULTIMEDIA INTERFACES
There is presently an extensive body of techniques that allows for the interfacing with
multimedia systems. The development of the use of windows, the application of high resolution
screens and the capability to transmit at Gbps rates allows for the implementation of much more
sophisticated technology. In this chapter, we focus on all the interfaces to the multimedia
elements, those that are direct to the end user and those that are more closely linked to the
processors.
In the last chapter we focused on the different types of presentation alternative afforded to the
end user and in the process developed sizing models for the different types of media objects. In
this chapter we focus on the human interaction elements with a multimedia terminal. This
interaction has two basic bounding elements. The first is the understanding and sophistication of
the end user and the second is the complexity of the application that is being used in this part of
the overall environment. Typically we desire that the end user have a minimal amount of training
for the system, however, this often goes contrary to the desire to have a fully flexible design for
the system. The second desire is to have all applications share a common context of use rather
than developing a visual and tactile dissonance with the user. Such a goal may be desirable but
doubtful ever reachable.
This in this chapter we develop and overall architectural paradigm for the interface to the
multimedia system and then give some examples of interfaces that exist today as examples. We
then extend the effort into the modeling of the interfaces and the modeling of the total end user
source, including the effects of the image characterization. This modeling effort is a critical
element in the analysis of the overall design of multimedia system since it helps answer the many
system and human factors questions that have been developed over time.
In the paper by Hartson and Hix, they clearly state that the objective of interface design is not
just to construct good interface but to develop environments in which good interfaces can be
generated. In this chapter, we focus not on the interface as an end in itself but as a means to
develop better interfaces. Thus our first efforts are directed at analyzing interfaces and their
performance. We specifically develop models for interfaces so that we may determine their
performance relative to a set of predefined performance criteria.
We then use that performance modeling as a tool to develop a synthesis of the Harston interface
design, knowing the methodology of analysis. Interface design is almost a Marxian process or
Hegelian dialectic, thesis or first attempts, antithesis or responses, and hopefully a synthesis of
154 | P a g e
the key elements of a good design. To reach that however, we need to understand what
constitutes a good design. Our approach is to focus on as many tangible issues that are possible.
4.1 INTERFACE ARCHITECTURES
Many works have been written on the human factors issues of man machine interfaces. The
simplest rule however is the one that states that the interface should be intuitive and lead to a
minimal amount of sensory dissonance. Sensory dissonance is the phenomenon that results when
the human user is asked to do something in a way that is radically different that he has been
accustomed to. A typical example of tactile dissonance is found in the area of operator service
stations in the telephone companies where calls are made to obtain telephone numbers. The key
entry devices are not in the standard QWERTY format but in an ABCDEF format. This was
done after it was found that the incoming operators could not be required to have typing skills.
The net result is that when an experienced person tries to use the system there is sever
dissonance and the response time is dramatically reduced.
The development of interface architectures may proceed along many lines. Sutcliffe has
described the Command Language Generation methodology with its four levels. These levels
are:
Task Level:The level at which we have input and output entries and we focus on what is being
done. Specifically this level focuses on the need that are satisfied by the user in applying the
interface mechanism.
Semantic Level: At this level we deal with the meaning of the processes that we are performing
and we deal with the underlying semantics of the tasks. Many elements of this approach deal
almost with the Chomsky like analysis of language and the ability of the language to convey
meaning.
Syntactic Level: This relates to the formal relationship of the formal interrelationships between
the symbols or words in the interface language or grammar. The syntactic level deals with the
semiotic elements that focus on the general philosophical theory of signs and symbols and their
use in natural languages.
Interaction Level: This is the simplest level in that is focuses on the simplest manual entries of
the end user. What key movements are made and how is the mouse moved to enter the data.
Thus we must consider developing an architectural construct for the multimedia interface. This
construct should first start with a clear understanding of the requirements of the processes being
executed and the sophistication of the user executing them.
155 | P a g e
4.1.1 Requirements and Specifications
The design of interfaces requires and understanding of the application and the level of
expectation son the part of the end user. There are however several factors that are general in
nature and are common in all user interface requirements. These elements are as follows:
Stability: The interface operations should be stable in that actions on the part of the user do not
lead to fault states that are unrecoverable. A typical stable interface requirement is one that sates
that there shall be no set of inputs that can occur that would lead to a locking up of the system.
Recoverable: This system should have the capability of providing a resolvable mode of operation
that allows the end user the ability to return to where they came from without excess operations.
Typical in this area is the use of the ESC key in PC applications.
Consistency: This implies that the same result will occur if a transaction is entered in the same
way no matter what state the system is in at the time of the transaction entry. It means that there
is no level of ambiguity of response that should be anticipated by the end user.
Non-Ambiguous: The interfaces requests for responses should be clear and unambiguous and the
level of semantic complexity should be kept at a minimum. If two alternatives are presented to
the end user then both alternatives responses should be equally stated in the presentation.
Minimizable: This implies that the transitions needed to effect certain actions should be of
minimum duration. The steps necessary to effect any transaction should not be excessively long.
Extensible: The system should be such that all states are reachable\e from any other state,.
4.1.2 Elements and Alternatives
The architectural models of the interface that we are developing in this sections are composed of
several layers of elements. We can consider these to fall into two general categories. They are the
static and dynamic characteristics.
The static characteristic of the interface describes how the interface is viewed in a single
interaction. We can look at the static user interface to be composed of the following elements:
Input: This represents the method, means and technique to enter information into the system. It
can range for a mouse entry to a keyboard entry and include touch screen or even digital
scanning devices. The entry technique depends upon the specific application, the nature of the
end user and the environment.
156 | P a g e
A classic example of a total mismatch of input mechanisms is the attempt to put a mouse on the
floor of an investment baking house trading station. In that environment, and device that is not
tied down to withstand hurricane forces can and will become a projectile. Thus the first sets of
mouse type entry devices found their way flying across many a trading floor. It sounded like a
great idea for the UNIX designer but for the actual user was appropriate for other applications.
Output: This represents the means, methods and techniques to retrieve information from the
system. The output in the static environment represents the one time application of the
information transferred to the end user.
Presentation: The presentation element of the static design represents how the information is
presented in form and design to the user for the purpose of inputting data obtaining output or
moving into a more dynamic mode of operation.
The dynamic elements are as follows: States, Tasks, Dialogs, Transitions, Interactions
4.1.3 Performance Issues
In evaluating interface systems, as we stated earlier in this chapter, we are basically developing
models for the interface and showing how the interfaces relates to and performs within a larger
and overall multimedia communications environment. Specifically, the interface system should
ensure that a rapid and error free level of performance is attainable by the user.
Thus as part of the development of a multimedia source interface description, we need to

determine the following types of performance measures:
Source Rate: The source rate of the total interface can be generated from the combination of
source characterization of the complex images as well as the models of the source within the
context of the human interface.
Response Time: The time between requests and responses is termed the response time. Thus the
time that the user requests an image and the time that the image is delivered is one measure of
the response time. Another, more general abstraction, is the response time that is from the time
the user begins the inquiry to the time the inquiry is complete. This is more than the time from
the request of image to response of image. The image is only part of the inquiry process. We
must be capable of modeling and accounting for the entire inquiry process, including but not
limited to that of a specific multimedia image.
4.2 PRESENTATION INTERFACES
157 | P a g e
There has been a significant development of various presentation interfaces over the past few
years. This development has focused on the need for a set of standard and portable interfaces and
development environments that can be used for a wide variety of applications. The driver for this
has been in many cases the world of UNIX and C. UNIX is fundamentally an operating system
that was developed in response to the need for a multiuser/multitasking environment that could
function on a smaller size machine (PDP 7). C was the language that was developed to
implement C and satisfied the needs of the developers as an elegant character and string
manipulation medium. Neither was developed in the context of dealing with less sophisticated
end users and moreover neither envisioned the growth of sophisticated display and I/O devices.
Thus it has become necessary to develop overlays to these basic elements of the computing
world.
This section details several of the more common examples and presents them in the layered
architectural context that we had developed in the previous section.
4.2.1 X Windows
The X Window interface is a current example of the end user interface development capability
that is available for use on many platforms. X Windows was a joint development effort that was
done at MIT for the purpose of running on top of UNIX type operating system and affording
the C programmer a more effective end user interface capability. X Windows provides and
interface to the applications program and in connected by calls to the X Library of calls. The text
by Schleiffer, Geddys and x provides significant detail of the operations of the X interface.
X is en environment, an environment to develop the user interface. It is not itself a user interface.
It is a fairly complex environment that allows for the interfacing not only with the user but with
the applications program and the network of other users in the system as well as complex form of
multimedia information.
X is composed of the following sets of elements. They are:
Application: The application is the end user program that is intended for use by the user for
effecting several transactions on the system.
X Server: The X server is a program that is run on the users machine or another machine but it
acts as the intermediary between the applications program and the handling of input and output
to the display. The X Server is the key X ingredient as an operative agent in the communications
network.
Workstation: This is the physical device that an application is run on. Frequently in the X
environment, there is one X Server per workstation.
158 | P a g e
Clients: The clients are all the users on the X environment and may usually be represented by
other applications programs.
X Library: This is a set of functions presented in the form a X primitives that support the X
windows applications. The X Library )(X lib) is structured to support a specific C language
environment.
X protocol: This is a set of tools that allow a single user to provide other users interfaces to be
independent from other clients.
X Window Manager: This provides for the management facility for the window layout and
interaction and assists when many windows are in action.
X tool Kit: This is a user interface subroutine library that does certain complex tasks using
already written X code that employs the primitives from the X lib.
Events: These are time stamped results of users actions such as the entry of a key stroke or
movement of a mouse.
The overall resource architecture available in the X environment. This resource architecture
allows for the access of such elements as: Windows, Graphics Contexts, Fonts, Color Maps,
Pixel maps, Cursors.
A typical X systems architecture shows the relationship of the workstations and their related X
servers and shows how this flow into the network through the X protocol. Using the X lib
functionality, we see how X provide access to the other elements of the X architecture. The X
protocol provides for the control over the flow of 11 of the multimedia elements in the system.
There is a significant disadvantage of this layered architecture. It is that the X protocol can often
cause thrashing of the data elements as the system tries to display complex multimedia objects.
In the present configuration, it is not common for the X protocol calls to file servers to take 15
to 45 seconds to display a high resolution image.
There are many types of servers in an X environment. They range from printer servers, database
servers, name servers and communications servers. Each can be addressable within the X
context.
the X window environment has a hierarchy of the windows themselves. The hierarchy allows for
the definition of a parent window, shown as WO in this example. From this parent we can define
a course of other windows that can be generated and manipulated.
159 | P a g e
X lib is structured into the following major functions:
1. Display Functions
2. Opens displays
3. Places information onto the display
4. Closes the display
5. Window Functions
6. Create
7. Destroy
8. Map
9. Unmap
10. Attribute assign
11. Configure
12. Translate
13. Window Information Function
4.3 GRAPHICS USER INTERFACE (GUI)
We have seen that X windows and the more applications oriented support of DEC windows or
UIL allows for the development and support of the end user interface. This has been formalized
in the structure of the Graphics User Interface architecture (GUI) that shows how these elements
relate to the hardware, operating system and ultimate the applications and the end user.
The general architecture of the GUI contains the following elements:
Hardware: This is the standard platform on which the system will operate.
160 | P a g e
Operating System: The standard operating system of choice. Frequently this will be UNIX,
PS/2, VMS or DOS.
Graphics System: This element is the actual graphic system used for the generation of the display
graphics. X in X lib has a certain graphics system capability but is very limited as a user friendly
graphics generator. PHIGS, GKS or PostScript has the better graphics capability.
Window Environment: X is a typical window environment. The wind approach allows for the
opening and closing of specific applications oriented windows that allow for the focusing of the
users interest and application.
Applications Manager and File Manager: This is the applications and often platform specific
display interface that allows for the direct interaction between the screen and the window
manager.
User Interface: This is the direct end user interface. OSF Motif is a typical example.
Applications: This is the end user applications.
There are advantages and disadvantages with the types of use interface development tools that
we have just been describing. They are:
Advantages
Provides a modular and transportable development environment.
Disadvantages
Adds significant operations overhead on the system thus introducing delays and other operational
factors.
Does not allow for customization and real time optimization of the applications.
Requires significant memory allocations to store the library functional and other parts of the GUI
elements.
4.4 MM SERVICES
Having developed some of the issues related to the development environment of the end user
interface, we can now focus on several specific end user interfaces and review how they meet the
requirements and specifications that we have discussed. All of these interfaces have been
161 | P a g e
developed in an X windows context and thus have been developed within the development
context discussed in this section.
The session screens allow the end user to enter into sessions with other users, applications
programs, or data bases. What we have shown is a simple set of directions for accessing and
determining the specifics of the session operation.
4.4.1 State Machine Analysis and Petri Nets
We shall now take the state diagram construct developed in the last section and apply it to
several specific applications. Specifically we shall apply it directly to the four services that we
have discussed in the last major section; mail session, file and directory. These services are the
type that would be of use to the typical end user of the communications service.
Presentation: This represents one screen that presents the information to the end user. The
presentation element is the heart of the state dynamics model.
Input Action: This is the set of input actions that are made in response to the elements that are
presented on the presentation screen.
Read Data Base: This action is one half of what is called a database transaction. It reads a
database element that is used as an integral part of the overall end user effort.
Write Data Base; This is the second part of the total database transaction and represents the
second possible response to an end user input action.
Process: This is the actions that are taken internal to the system that combines the result from the
presentation, that actions of the database transaction and the internal mechanism that are part of
the user interface.
Output Action: This is the action that results from the process and the database transaction
elements or it may even result directly from the input action and the presentation. The output
action then leads to movement to another presentation.
Using the state dynamics model, we can see that the states are effectively the presentation
elements, and that the transitions between are comprised of all, of the other elements. This
transition path is a block representing the transition process that we have developed in the state
dynamics model.
162 | P a g e
Consider the example of a simple user interface that is part of the withdrawal of funds from an
ATM. We have four presentation elements or screen that are presented to the end user for s
decision to be made. The screens are:
Screen 1: Indicates that the user must log onto the system with some PIN number or other
security code mechanism.
Screen 2: Asks the user if they desire to deposit money or withdraw funds.
Screen 3: This is the withdrawal screen and has all the information necessary to withdraw funds.
Screen 4: This is the deposit screen and it also contains all the necessary information.
Movement occurs between the stets. The movement requires that the other elements of the model
be evoked, namely the database elements, and the input and output functions, each associated
with a particular screen. For example, let us focus on screen 3 which is the withdrawal screen. It
has the following elements:
RDB
Presentation Input Action WDB Output Action
Process
Presentation
Presentation
Presentation
Presentation
163 | P a g e
RDB
Presentation Input Action WDB Output Action
Process
Sk Sn
Sk Tj Sn
In this figure, we show that the basis element are the states and the transitions. These two
elements can be combined into a general theoretical structure called the Petri Net. We shall rely
upon the work of Marsan et al to develop this theory. This we shall also extend with the work of
Stotts and Furata which combines this with the Hypertext model. We can now begin to define the
Petri Net.
Definition: The Petri Net, PN, is a tuple consisting of the following elements:
P = A set of places = {Pl,....,Pn}
T = A set of Transitions = {Tl,....,Tm}
A = A set of directed arcs; (PXT)X(TXP)
and :
PN = {P,T,A}
In a PN there are two sub sets call pre sets and post set. These are the elements that make up the
arcs. We define these as follows:
preT = {P: (P,T) F}
164 | P a g e
and
postT = {P: (P,T) F}
where F is the set of all acceptable P,T pairs.
Consider a sample Petri Net as shown.
p1
T4 T1
p4 T5 p2
T3 T2
p3
Using the state dynamics model, we can see that the states are effectively the presentation
elements, and that the transitions between are comprised of all, of the other elements as shown.
In this figure we show the relationship to the state and the transition path. This transition path is
a block representing the transition process that we have developed in the state dynamics model.
In this figure, we show that the basis element are the states and the transitions. These two
elements can be combined into a general theoretical structure called the Petri Net. We shall rely
165 | P a g e
upon the work of Marsan et al to develop this theory. This we shall also extend with the work of
Stotts and Furata which combines this with the Hypertext model.
We can now begin to define the Petri Net.
Definition: The Petri Net, PN, is a tuple consisting of the following elements:
P = A set of places = {Pl,....,Pn}
T = A set of Transitions = {Tl,....,Tra}
A = A set of directed arcs; (PXT)X(TXP)
and :
PN = {P,T,A}
In a PN there are two sub sets call pre sets and post set. These are the elements that make up the
arcs. We define these as follows:
preT = {P: (P,T) F}
and
postT = {P: (P,T) F} where F is the set of all acceptable P,T pairs.
Consider a sample Petri Net. In this example there are four places, similar to states or
presentations in the original model description. There are five transitions. These transitions
represent the accumulation of all the steps that are required to be accomplished to get from one
place to another. The transitions control the movement from one place to another, whereas the
places are less active players in the PN model.
We have also identified the pre and post sets by the use of the terms aij, where we use a pre and
post notation attached.
We can now introduce the concept of a marking, which is a way in which movement can bet
determined around the PN> A marking is simple a set of integers that are assigned to a set of
places, and an algorithm that details how those integers are propagated around the network. Let
us define a marking as follows;
166 | P a g e
Definition: A marking is an n tuple, {bl,/bn}, where bk is a binary integer, 0,1, and;
M = {b1….bn}
where;
M: P-> {b1….bn}
Further we define MO as the initial marking and Mf the final marking.
Firing of a PN consists of taking one token in each preT and adding it to the token in postT.
we can now further define the actual execution algorithm for a PN.
Definition: The execution algorithm for a PN consists of the following steps:
(i) Any Ti is enabled when ALL of its input places contain one or more tokens.
(ii) A transition Ti, that is enabled; can fire and when it fires it removes one token from each
input Pk and places a single token in each output Pj.
(iii) Once the firing occurs, the tokens are repositioned and a new marking occurs. The marking
sequence may be given by M(0), M(l), , M(n),...
It should be noted that there are two non-determinitsic token transitions that can occur. These are
at the deposit/withdraw place and at the Another Transaction? place. In this case we shall leave
them deterministic. In latter parts of this chapter we shall sow how we handle this
probabilistically.
167 | P a g e
AMT T
1
Chk
D
T1
T1
T T Anot
Start 1
PIN 1
D/A
her
T1
T1
AMT T
1
Cash
W
m1 m2 m3 m4 m5 m6 m7 m8
1 0 0 1 1 0 1 1
1
0
0
1
0
In the transition of these states, we see that M(0) is the initial state and that initial state may be
anything that we wish to start with. From an initial M(0), we generate the sequence of markings,
M(l),....,M(n),... These are generated from a generic state transition function;
M(l) = G(1,0) M(0)
or in general we have;
M(k+1) = G(k+l,k) M(k)
The transition function can be calculated based on the transition elements and the arcs. We leave
this detail to the problems at the end of the chapter.
168 | P a g e
Having defined the concept of the PN we can now extend this to the concept of a Hypertext. This
is developed in the context of the work of Stotts and Furuta. It is as follows;
Definition: A hypertext is an n-tuple that is as follows;
H = {PN,C,W,B,Pl,Pd} where;
1. PN = any Petri Net
2. C = a set of document contents. This may include any set of text, graphics etc that may
make up the hypertext document.
3. W = a set of windows. These windows may be ordered or otherwise.
4. B = a set of buttons. These are any actions that may create a response on the system.
5. PI = a logical projection of the document.
6. Pd = a display projection of the document.
Using the concept of the PN Hypertext we can determine several important solutions to problems
concerning Hypertext. These problems are:
1. Display Complexity: Using the PN graph and the Hypertext adjunct, it is possible to
determine the number of hypertext windows that are simultaneously need to display the
information. This is the number of marked places in the PN associated with the
Hypertext.
2. Path Size and Synchronization: Using the PN formulation, we can determine the length of
the path needed and the level of concurrent path synchronization.
3. Reachability of Places: This is performed in developing the set R(M(0)), the reachability
states of the initial state.
There are several extensions to the Petri Net concepts that we shall develop. These will become
essential as we develop the model for the sizing of the source characteristics.
Definition: A PN place Pn has multiple arcs. The threshold Tj associated with Pk fires if and
only if there are two tokens in Pk. In addition, only one token is passed onto Pn.
169 | P a g e
Multiple Tokens
Start T1 PIN
ID
Inhibition
T1 Action
PIN
Definition: An arc is called an inhibited arc, if and only if all the arcs contain markings, except
for the inhibitor arc.
Definition: If we let R(MO) be the reachability set of PN, with initial state MO, then the dead
markings are those sets in R(MO) that go nowhere. Namely;
M(DS) is a dead set iff;
MDS(k+l) = G(k+l,k) MDS(k)
Definition: PN is called a safe net if the number of tokens <= 1 for all Pj and for all R(MO). We
further define PN as STRICTLY CONSERVATIVE (SC) if for all M in R(M(0));
sum of tokens is constant
If PN is SC, then if C(x) is the cardinality of the set x, we have:
∑ C(preA) C(preA) = ∑ C(postA) i=l i=n

170 | P a g e
We now introduce two new concepts. The first is the timed PN and the second is the stochastic
PN. The timed PN introduces the concept of timing for the PN concept. The stochastic PN allows
for the generation of PN transitions that are stochastic in nature. Specifically we have introduced
the concept of random timing that is the basic element in determining the ultimate source
characteristics.
Definition: A timed Petri Net, TPN, is an n-tuple where D is a set of delays at each of the
transitions before firing. Specifically, if we look at the example, we see that at transition Tk, we
have a delay Dk that is in place before the transition fires.
P4
T1 T2 T3
P1 D1
P2 D2
P3 D3
Let us expand on the concept of the stochastic PN. In this case we expand on the TPN by
applying the random variable to the timing at the transitions.
Definition: A stochastic Petri Net, SPN, is an n tuple that is:
SPN = {P,T,A,M(0),L}
where L is a set of random delays at the individual transitions.
We can then define the probabilities associated with the system going from one state to another,
or even the probability of being in any state. In particular it can be shown that the SPN is
isomorphic to a Markov chain.
Finally, we introduce the structure that we begin this section with. Specifically, we introduce the
concept of random transition path selection, as well as random delays at transitions. We also
171 | P a g e
introduce the input and output of data at the time of each transition. We first introduce a simple
example to develop this concept.
Example: Consider the sample system that depicts pictures, text, or video. At two nodes we have
options that could occur, one allow for the choice of the three images, the second is to decide if
the next step is logging off or of starting over again.
T Imag T
1 e 1
LogO T T T
1
Select 1
Text 1
Next
n
T T
1
Video 1
What we see in this SPN is that there are non-deterministic choices for these two transition
nodes. To alleviate these non-deterministic problems, we introduce an random selection choice at
each node. Thus we can define the probability of choosing video as ql and that of text as q2 and
that of video as q3. We note that the sum of the q's is unity as expected.
In addition, when we decide to see a picture, when the transition is activated, it generates an
output that is sent to another similar machine. The output is a packet of information of certain
length. In a similar fashion, there is an input packet that contains the information necessary to
regenerate the picture requested.
Definition: A Source Dynamic Model (SDM) is an n-tuple that is characterized as follows:
SDM = {P,T,A,M(0),L,Q,I,0}
where the first part of the SDM is a SPN, and;
Q are the transition node probabilities
172 | P a g e
I are the inputs that occur at each transition when activated
0 are the outputs that occur at each transition when activated.
We shall use this model for the development of the source characterization.
4.4.2 Performance Analysis
The modeling of the multimedia source will require the use of probabilistic techniques that are
found in many other areas. In particular we shall show that the end user interface state diagram
ca n be modeled effectively using the finite state Markov state machine model and that statistics
on the overall state averages can be readily drawn from the technique.
We shall be focusing on the development of performance issues as well with the development of
the source model. These typical performance issues are:
1. Response Time
2. Average Holding Time
3. Number of Transitions per Unit time
4. Stability of the state protocol
5. Source generation rate
4.5 MM SOURCE MODELING
Multimedia source modeling combines the factors developed in the interface model and
combines them with the model development that we performed in the last chapter. We can
therefore takes a session interaction at one user, and develop the types of input to a
communications network that would be anticipated from such as system source. We can then
further expand this source to include a collection of sources to encompass a complete array of
users that can provide the cumulative load on the total multimedia communications network.
The source modeling will include the combining of the elements developed in the system state
diagrams with the additional factor of image sizing and packaging within the dialogue of the end
user session.
4.5.1 Model Elements
173 | P a g e
It contains the elements that we have developed in the SDM system. It is based on a standard
Petri Net models and we have introduced the timed transitions and the stochastic selection of
transitions at the places indicated.
T
P6 1
T
T1 P3 1
T
P1 1
P2 P5
T
T1 P4 1
T1
P7
To develop the source model, we need to do the following steps:
1. Define and develop the PN for the characterization of the Hypertext interaction with the end
user.
2. Assign probability distributions to the firings of the transitions. Typically, exponentially

distributed random variable are adequate for this assignment.
3. Assign transition probabilities to each set of non-deterministic transition. A non-deterministic

transition is one in which an arc from a place to a transition occurs several times from one place.
The transition probabilities are for each non-deterministic place, Pj, and we have the transition
from Pj to Pk or Pj to Pn. Define the transitions:
q jk = P[to Pk | from Pj]
and note that for each j:
qjk =1
k Hj
174 | P a g e
where Hj is the set of all transition places from Pj. 3.4.2 Model Structures
The source model characterized in the previous section combines the results that we had
developed in and those that we have developed in this chapter. In particular, we can see that
the source can be given in terms of a fully stochastic model. Let us begin by modeling the input
and output responses for the SDM. These responses are the corresponding input and output
traffic on the communications network.
The input and output messages consist of the following possible elements:
1. Images
2. Stills
3. Video
4. Voice
5. Text
6. Graphics
We can now characterize each of these in terms of bits of information, and furthermore
characterize them in terms of data rates, namely bits per second.
Since the SDM is a stochastic Petri net, which is in turn a timed Petri net, we know that
responses occur a set of define, albeit random, times. We call these times {LI,...,Ln,..}. Thus we
have a sequence of input and output messages. Let us define Mi(t-Lk) as the input message at
time Lk and Mo(t-Lj) as the output message at time Lj.
There are two classes of sources that we have developed. They are continuous and transient. A
continuous source is one that never stops and continuously generates input and output messages.
A transient source is one that does not last indefinitely and has a continuously decreasing
probability of Lm continuing. We can provide a more definitive definition as follows.
Definition: A place Pterm is called a terminal piece if no transitions emanate from Pterm and
there are no input or output messages associated with Pterm.
Definition: A SPM is said to be continuous if there is no Pterm.
Definition: A SPM is said to be transient if there exists a Pterm and if for all M(0) Pterm belongs
to R(M(0)), and M(k) converges to Pterm as k increases. Specifically;
Let M( ) = {x,x,, z}
175 | P a g e
where z marks Pterm and can take 0,1. Let:
M(oo) = {0,0,,0,1}
Then M(k) approaches M(oo) with probability approaching 1 as k approaches oo.
We can define the input and output message streams as follows: mout(t)
= mo,1(t-Ll)+mo,2(t-L2) ++mo,k(t-Lk) +..+mo,n(t-Ln)
min(t)
= mi,1(t-Ll)+mi,2(t-L2)+...+
where:
mo,k(t-Lk) = response, request, image etc.
which depends on the probability assigned or directly upon the place transition characterization.
3.4.3 Model Performance Analysis
We can now take the models that we have developed and convert them onto source models for
the purpose of systems performance analysis. PI is the initial state and P4 is the final state. We
associate the time delays Lk with each transition.
We start by enumerating all of the paths through the network. A path is an ordered sequence of
places that define the network. Associate with each path is a probability of occurrence. We
assign to the nondeterministic choice at P3 by assigning the probability of p to going from P3 to
P4 and q or (1-p) to going again to PI. With this we can assign the following:
Path Probability
P1,P2,P3,P4 p
P1,P2,P3,P1,P2/P3,P4 p*q
PI,P2,P3,PI,P2,P3,PI,P2,P3,P4 p*q*q
176 | P a g e
etc. We can readily add the probabilities and see that it adds to unity. Associated with each path
is a duration Dk, where:
Duration Probability
L1+L2+L3 p
L1+L2+L3+L4+L1+L2+L3 pq
etc. From this we can determine the average duration of a transient SDM.
We define E[L] as the average path length in seconds, where it is the length of each path times
the probability of that path.
In a similar fashion, we can generate a model for the total number of input and output bits
generated in this network. We do this again by following the progress through each of the paths
in
the network. We can then define the average number of input and output bits as;
E[I] = average input bits
E[0] = average output bits
Using this and the source duration we have the average input and output rates of the source;
Ri = E[I]/E[L]
Ro = E[0]/E[L]
If we have many such sources, then we can generate an average source rate by knowing the total
number of sources, determining how frequently they are activated, and then averaging this over
the individual sources own statistic. We expand on this in the problems.
4.6 CONCLUSIONS
This chapter provide the designer with the second element that is necessary for the development
of the source model in a multimedia context. It allows for the characterization and development
of a multimedia stochastic source and the specification of that source in terms its average
characteristics as well as a full probabilities characterization. We shall see how this concept will
be latter integrated into the overall system design and evaluation.
177 | P a g e
The key issues to be obtained from this discussion are those that relate to both the design and
evaluation of the end user interface. Many works have been written on the end user design
factors but as we continually see there are all too often designs that do not provide for ease of use
and simplicity of access. There is the continual battle to provide the end user with all the
flexibility that could be desired while at the same time increasing the complexity of the interface.
178 | P a g e
5 MULTIMEDIA STORAGE
The storage of multimedia objects is a major issue not only to the basic archiving of the data but
goes to the heart of the issue of multimedia characterization. As we have discussed before, the
ability to do more with complex multimedia objects than just raw digitization will represent the
major capability to deal with complex abstractions. Thee complex image abstraction concept is
the heart of multimedia storage.
This chapter discusses the ways in which complex multimedia elements can be compressed and
reduced to simple data elements. It also addresses the more complex problem of how best to
abstract elements of images and to combine them in a fully interactive multimedia session.
Storage is one of the elements that are key to the overall design and performance of the
multimedia communications environment. Often the concern is the availability of
communications channels that are fast enough to bring the images to the end user. The problem
in reality is not the communications channel but the storage element. In this chapter, we focus on
the individual storage elements and blend them together into a full multimedia storage capability.
In addition we focus on the system issue associated with the system performance and sizing that
is the driving force in all of the design tradeoffs in this text.
5.1 STORAGE ARCHITECTURES
In this section we present an overview of the key characteristics of storage devices and in
addition provide the key factors that go into the performance analysis of the storage systems.
Storage Elements
Storage Access Characteristics
Storage Performance Parameters
The overall performance factor for the operations of a storage system is the access time for a
specific file element. In a multimedia system, the performance factor is the access time for the
compound multimedia file element. There are four major elements of the access time. These are:
Bus access speed and time
Memory unit latency access time
179 | P a g e
Data element seek time to find the data.
The data transfer time from the medium to the memory display device.
Multiple Storage Techniques
We have just completed the analysis of system that have single storage media and the
performance that results. When dealing with
multiple media we are dealing in two dimensional of further integration. These dimensions are:
Image context integration, integrating voice, video, text, images etc from differing multimedia
sources.
Storage device type integration, taking multimedia storage from multiple devices at the same
time and creating a compound multimedia storage construct.
5.2 STORAGE ALTERNATIVES
There are many storage media that are available for the temporary storage of the multimedia
elements that are used in an operational context. These media are in range for very fast but not
grate in extent, to fairly slow but of high density. One of the driving factors for the choice of the
specific storage medium is the cost per bit of storage and its accompanying access time. In this
section, we shall not focus on the cost issues since they are generally so volatile but urge the
reader to review the issue whenever developing a system design.
High Speed Memory (RAM)
Disk Storage
CD ROM Storage
Tape Storage
5.3 FILE FORMATS
Having the specific storage device is one of the elements in the overall storage design problem.
The second, and equally important, is having the proper file formats. There are many types of
file formats for the efficient storage and retrieval of multimedia data elements. Clearly, it should
180 | P a g e
be obvious that the storage of voice, video and images may require different file formats from
both the perspective of placing the information in storage and in ultimately retrieving it.
5.4 MULTIMEDIA FILE RETRIEVAL ALTERNATIVES
Having discussed the overall issues, technology and file structure, we now will develop the
overall system view of a multimedia file receival system. In this section we shall focus on the
overall implementation of the architectural alternatives and develop methods and models for the
analysis of performance and sizing of the multimedia file structures.
Storage Devices
Access Techniques
5.4.1 Performance Measures
5.4.2 Access Optimization
Having developed the measures of overall, system performance, there are several issues that
relate to the optimization of the file system architectures. We have noted that several issues
aware key to the design and performance of a multimedia system. These issues are:
Access time per element
Multimedia access time and compound access integration time
Consistent file naming and formatting of data.
5.5 CONCLUSIONS
In this chapter we have focused on the element of the multimedia communications system that
stores and retrieves the data elements. The key issues are those that relate to the technological
alternatives and the concerns relating to the performance of the system as an overall
communications system. The issues of storage capacity, the layout of the specific files and the
analysis of the access time to retrieve files are key to the development of an efficient overall
system. We shall use these result as an element of the overall system design latter in the text.
181 | P a g e
182 | P a g e
6 NETWORKS AND MULTIMEDIA
The telecommunications network has evolved from a structure that was initially imbedded in a
regulated monopoly through one that has been dictated by judicial mandate. The natural
economic forces that tend to work in other market segments have been deliberately left out of
this market. The result is a fragmented and less than efficient market for the delivery,
development and expansion of services to the end users. There are however, alternatives that
may allow for a repositioning of these current structures and permit a restructuring of the current
communications market infrastructure. This paper develops a set of alternative architectural
constructs, integrates several technological trends, and describes multiple evolutionary paths.
The structure and viability of the paths are discussed in terms of their economic viability. This
paper also discusses the potential and evolution of broadband networks and their role as
infrastructure elements in a national network.
6.1 INTRODUCTION
Networks have been developing in various forms since the time of divestiture, in response both
to the opportunities afforded by deregulation as well as by the needs of the users themselves.
There has evolved a clear lack of cohesiveness to the design and application of the network
schemes and the evolution seems to presage a movement towards networks optimized for usage
by a defined and bonded collection of shared users. The National Research and Education
Network (NREN) concept clearly falls into this category. To better understand the implications
of such evolutionary trends and to better develop a base of knowledge for the development of
effective policy in this area, it is necessary to have a construct or model for these evolving
networks.
The NREN concept is an evolutionary progression of networking capabilities that starts with the
introduction of a shared data network at the 1.5 Mbps rate and then transitions to a 45 Mbps set
of rates and then in its third stage it acquires a Gbps capability. The major focus is on that later
stage capability.
It is the development of this area that has been considered as the infrastructure portion of the
effort. To quote from an OSTP report, (See Kahin, p.4);
" The NREN should be the prototype of a new national information infrastructure which could be
available to every home, office, and factory.”
This raises the expectations from that of a network to that of an infrastructure. In this paper, we
shall develop the concept of infrastructure against the context of communications networks. We
183 | P a g e
shall address the issue of NREN and its counterparts being either infrastructure or merely another
network. For in the same paper, Kahin goes on to define the NREN as;
" Despite the name, NREN is not conceived as a centralized national network...The NREN is
conceived as more coordinated than the present Internet...the vision of the NREN generally
includes eventual transition to commercial users.”
Thus there are two views of the NREN; that of a prototype network and that of a fully
operational entity.
There are similarly multiple views of many of the types of networks that may be developed for
the purpose of developing national expertise in the areas of broadband communications. The
views, as indicated, flow from those that view the need for a purely research oriented network to
those who see the current need for a fully operational physical national infrastructure.
This paper addresses several of the fundamental issues that may assist in resolving the issues
presented by these network alternatives. Specifically, we address the concept of infrastructure,
and describe its multiple embodiments. We then develop the concept of architecture and how
that it includes several elements and that a network can be viewed only in the context of its
architectural embodiment. We then develop the major exogenous drivers for any network, the
end users. Finally, we consider the elements of policy and how policy may be developed in the
context of the evolving network world views.
There are several questions that we investigate in this paper. First, what is a network and what is
a network infrastructure, and within that context, how does one create a broadband, high data
rate, network infrastructure. Second, does such a network capability already exist and if so how
does one assess and use it as is. Third, what are the goals that are achievable with such an
infrastructure network, in terms of international competitiveness, establishing an information
infrastructure, or in terms of a national asset.
The main theme of this paper builds on these questions. Specifically, it will be argued that there
are basic architectural alternatives for network designs. The existence of these different
architectures is based upon a world view and a set of technologies that enable its embodiment.
We argue that there is a fundamental change in the world view, one from a truly hierarchical
environment to one that now empowers the end user. The result of this change is a fundamental
change in the operative network architectures, moving from the world of large scale
infrastructures to those of multiple overlay networks.
The current dominant carriers in the telecommunications market are generally regulated on the
theory that they have monopolistic power in their respective markets by either a direct or implied
184 | P a g e
exclusive franchise to provide services to the customers in those markets. Furthermore, market
control allows the dominant carrier to exert price control as either a monopolist in the pure sense
or at least as a simple oligopolist. It is the concept of the "Bottle Neck" that has dominated not
only policy but also the flow of technology and services to the market. In this paper, we
demonstrate that the dominant carrier concept is degenerating in many areas. It is becoming an
environment of multiple network providers, where the functions of the classical carriers are now
becoming distributed directly to the end users. Furthermore, we shall argue that as a result of the
technological changes allowing for multiple carriers, that the current regulatory strictures on the
telecommunications market are not only inhibiting to development but are the fundamental
essence of the loss of international competitiveness.
The policy issues associated with these classical systems are based upon the fact that, in the past,
there has been a perceived lack of growth and innovation, a set of barriers to entry on the part of
new innovators and price competitors, and a set of price structures that inhibited the entry of
competitors. We demonstrate in this paper that there are now a multiple set of alternatives for the
network user. The alternatives are based upon a revised view of what a network is as well as
major changes in the technology of networking.
Thus, these changes require an expanded development of new policies for these network
infrastructures.
Networks have been viewed by some as an infrastructure such as highways, educational systems,
and the military. Typically they have been thought of that way because they serve the general
good, they require significant capital investments that are frequently beyond the bounds of most
single users, even corporations, and because they must deal with both national and international
elements of our society.
Thus, the three reasons for the infrastructure view are the general good, capital intensiveness and
global in scope. In this paper, we argue that the concept of networks and infrastructure must be
revised and expanded. That networks such as the NREN, are only one of several embodiments of
a network, and that infrastructure must be understood in the context of the new world views of
networks. Further, we argue that technology is changing what we can do with the network
architectural elements and that this added use flexibility to create subnetworks that are optimized
to meet their specific economic driving forces, makes the very concept of a single physical
network architecture obsolete.
Specifically, market demands require that the networks have more customized structures and that
in fact the general network that meets everyone's needs is both counterproductive and adds
additional costs.
185 | P a g e
Second, technology allows for segmentation in the network, using a commodity infrastructure of
fiber to allow the end users to have the use specific designs that they need. For example, if we
allow fiber to be used at its full potential to the end user, rather than as a segmented and
compartmentalized structure matching a voice only world, this may free both market and
technical forces to attain the segmentation.
Third, globalization has already occurred with the globalization of the business market.
There are currently five players in the networking area. They are the Government and its own
networks, Public Switched networks provided by the common carriers, Private Networks,
frequently called Bypass carriers, CATV networks, primarily carrying entertainment and not
information, and customer specific networks, such as those of IBM, DEC and other large
information intensive companies, as well as the regional educational networks, such as
NEARNet and NYSERNet. The five network providers and generally depicts the gross
differences between each of them. It is essential to understand not only the microstructure of
each of the players in this map, but to also understand that each of the players has a dramatically
different world view of what they are to achieve in their networks. In this paper, we shall
examine each of these players in the context of the new world view of networks In this paper, we
will prove the following thesis:
The use of a centrally conceived, non-market generated and driven, means to develop an
economically productive implementation of a new electronic broadband communications
network is not only an ineffective act but also is counterproductive to the national economy.
We prove both sides of this thesis in the body of this paper. Moreover, we demonstrate that the
broadband network is and must continue to be driven by end users motivated by such basic
economic drivers as value creation. The evolution includes not only the basic transport features
but the entire complement of elements necessary for a viable communications and business
entity. The current environment is such that competitive market forces exist, inhibited only by
regulation mired in an obsolete world view that must be adopted to the new paradigms of
implementation.
The process that we take to prove this thesis is as follows:
First, we develop the concept of architecture and show that there is both a philosophical and
material structure to the concept of architecture. The key observation of this first step will be to
note that our understanding of communications architectures is dominated by a world view
created by the available paradigms or technologies.
186 | P a g e
Second, we demonstrate that the underlying technologies have changed dramatically in the area
of broadband. The change is dominated by the ability of the end user terminal to play a dramatic
role in the network operations and that ability to have a fully distributed design. This technology
change is reflected in new paradigms and thus agues for the development of a new world view.
Third, we demonstrate that the market is the ultimate arbitrator. That value creation as a
measurable and definable entity is critical to the success of any broadband implementation and
that this value creation must be perceived by and understood by the customer. Without the
customer as user, buyer and decision maker being the focal element of the process, the effort is
doomed to failure.
Fourth, there are several players all trying to evolve in the direction of developing the basic
elements necessary for a broadband environment. We review them in some detail and show that
there already exists many of the elements in place for such a system and that another scheme will
result in reducing the current competitiveness of these players and impact negatively on the
economy as a whole These four steps build to the proof of the thesis.
6.2 ARCHITECTURES
The concept of a telecommunications architecture has been a cornerstone in the development of

new telecommunications systems. However, the structural elements of these architectures have
not played a role in the development of policies. In this section we will develop the concept of an
architecture as a means to understand the network as both a market and regulatory entity, and
187 | P a g e
will provide a new set of perspectives for viewing the network in terms of a new paradigms and
world views.
An architecture, first, requires that the underlying system be treated in terms of a set of
commonly understood elements and that these elements have a clearly demarcated set of
functions and interfaces that allow for the combining of the basic set of elements. The way the
elements then can be combined, reflected against the ultimate types of services provided,
determine the architecture.
An architecture, secondly, is driven by two factors; technology and world view. Technology
places bounds on what is achievable, however those bounds are typically well beyond the limits
that are self-imposed by the designer or architect in their view of the user in their world. This
concept of architecture and the use of design elements is critical in understanding the paradigms
used in the structure of information systems (See Winograd and Flores, pp 34-50, especially their
discussion of Heidegger and Throwness in terms of design). World view is the more powerful
driver in architecture (See Kuhn, pp 72-85). We argue in this paper that it is essential to develop
a philosophical perspective and understanding of how to view networks. We argue with
Winograd and Flores, and in turn with Heidegger, that we must be thrown into the network, to
understand the needs of the users, and to understand the structure of the paradigms that are used
to construct the world view.
To better understand the importance of an architecture we develop the concept of the historicity
of architectures based upon the work of Kuhn and ten that of McLuhan. Kuhn begins his thesis
of how scientific revolutions occur by the introduction of the concept of paradigms. He defines
188 | P a g e
these as (see Kuhn p. 175);"...the term paradigm is used in two different senses. On the one hand,
it stands for the entire constellation of beliefs, values, techniques, and so on shared by the
members of a given community. On the other, it denotes one sort of element in that constellation,
the concrete puzzle-solutions which, employed as models or examples, can replace explicit rules
as a basis for the remaining puzzles of normal science, The first sense of the term, call it
sociological, ..., “
hand for the network and the ways we put these elements together. New paradigms result from
new technologies. New technologies allow for the placing of the elements together in new ways.
Kuhn, then goes on to demonstrate that the world view, that is how we view ourselves and our
environment is based upon the our acceptance of these paradigms, as either collections of
techniques and technologies or as collections of embodiments of these techniques and
technologies in "examples". We then end to accept this as the way things are and should be. Then
Kuhn argues, as the technologies change, changes in the paradigms do not occur in a continuous
fashion but almost in quantum leaps. The new paradigms build and congeal until they burst forth
with new world views. It is this model that we ague applies to the evolution of broadband.
It is this philosophical view, almost Hegelian in form, that is essential in understanding the
underlying and formative changes in paradigms that will change our world view.
As a second perspective of the impact of technology as a dominant driver, we can refer to

McLuhan and his development of the concept of media. Drucker has referred to the presentation
of McLuhan's doctoral thesis and McLuhan is quoted as follows (See Drucker, p. 250):
"Movable type, rather than Petrarch, Copernicus, or Columbus was the creator of the modern
world view.. "Did I hear you right," asked one of the professors as McLuhan had finished
reading, "that you think printing influenced the course the universities taught and the role of the
university, altogether?”
"No, sir, " said McLuhan, "it did not influence; printing determined both, indeed, printing
determined henceforth what was going to be considered knowledge.”
This concept later evolved into the medium being the message. In our context it is the fact that
both Kuhn and McLuhan recognized, albeit in differing fields and in differing ways, that
fundamental changes in technology and technique, call it paradigm or the medium, will change
the world view, also the message.
It is the importance of understanding the change in the technology, its function and evaluate the
possible change that this will have in the world view. It will be argued, that much of the thinking
189 | P a g e
in the current broadband areas, NREN in particular, is based upon outmoded techniques and
structures, and that a differing world view will evolve.
Thus, architecture is the combination of three parts; the common elements, the underlying
technology and the world view. The conceptualization of architecture as the amalgam of these
three elements. We shall develop this construct more fully as we proceed.
The concept of a world view is an overlying concept that goes to the heart of the arguments made
in this paper. To better understand what it implies, we further examine several common views
and analyze the implications of each. If we view our world as hierarchical, then the network may
very well reflect that view. If we further add to that view a bias towards voice communications,
these two element will be reflected in all that we do. The very observations that we make about
our environment and the needs of the users will be reflected against that view. As an external
observer, we at best can deconstruct the view and using the abilities of the hermeneutic observer,
determine the intent of the builder of the networks. (See Gadamer's interpretation as discussed by
Winograd and Flores, pp 27-30. Also see the historical context of the hermeneutic approach in
the sciences as discussed by Greene in Depew and Weber, pp 9-10).
Take, for example, the use of twisted pair, pairs of copper wire, to transport telephone traffic. For
years it was implicitly assumed that this transport medium was limited to 4,000 Hz of bandwidth,
that necessary for an adequate quality voice signal. Specifically the world view was that of a
voice network that was to be used for voice traffic only. Ten years ago, this was a true limitation,
since the transmission was forcefully limited to 4,000 Hz by inductive loads or coils on the
telephone lines, assuring that you could do no more than the 4,000 Hz of bandwidth. Then, there
was a short period in the mid-1980s, when Local Area network manufacturers found that you
could transmit 1.544 Mbps over the common twisted pair, and that data was viable in what was
assumed to be a voice only medium. What had been almost religiously believed to be a limit was
found to be untrue. Then with the introduction of digital switches, the old "inductive loads" were
returned with the switch now limiting the data to 4 KHz or 64 K samples per second. The world
view of a voice only network took hold again, but this time in the context of a data rate
limitation, rather than a bandwidth limitation. In the early 90's there is another attempted break
out of the world view and to put 100 Mbps on twisted pair, so called FDDI circuits. Again, due
to the limitations on the part of the network as a voice dominated system, the world view keeps
this high data rate capability on the customer's premise only, and not the network.
190 | P a g e
Here we indicate the two dimensions of information transport, bandwidth and data rate. The
designer of the transport facility may limit the data rate by selection of signaling format or
delimit bandwidth by filtering. It encompasses a large capability of either providing bandwidth
or data rates to the user. The two limiting world views are indicated as two solid lines, one at
4,000 Hz and one at 64 Kbps. Both are voice only world views. We can readily see, that with
optical fiber superimposed the same issue of architecture dominated by world view may result.
In the fiber case, the result may be a segmenting of the architecture along selected data rate lines,
again formed by the voice world view.
Thus, architecture can be defined as the conceptual embodiment of a world view, using the
commonly understood set of constructural elements, based upon the available set of
technologies. For example, Gothic architecture was a reflection of the ultimate salvation in God
in the afterlife, in a building having a roof, walls, floors, and windows, and made of stone and
glass. Romantic architecture was, in contrast, a celebration of man, using the same elements, but
some employing a few more building materials. The impact of the differences in world view are
self-evident in the embodiments of the architecture. (See the discussions on the impact of world
view on architecture in Wolfe. In addition see the cultural or world view impact on the Gothic
architectures in Jantzen and in Toy.) Let us consider a second example of the impact of world
view on architecture, specifically the difference between the ISDN architecture and the
architecture embodied in Local Area Networks, LANs. ISDN is an architecture consistent with a
voice dominated, hierarchical world view of single points of control. LANs are architectures of
world views that reflect both end user self-empowerment and the environment of a data driven
utility. The LAN embodiment as well as its extension in the CATV architecture of voice
191 | P a g e
communications using a LAN world view. This evolution in thought is critical to understand the
impact of world view. The LAN is an embodiment of empowerment of the individual view,
developed in the context of the 1960's and 1970's. The LAN concept, originating at such
locations as XEROX PARC, was driven by the developers needs to enable and empower the end
user with computing capabilities heretofore unavailable. Out of this view came the LAN
architecture of a fully distributed system, using a coaxial transport mechanism to do nothing
more than provide bandwidth. The transport mechanism is a broad enabler. The actual
implementation of the details is done at the users terminal in hardware and software. This is in
sharp contrast to ISDN, where the ISDN central switch does the enabling. In ISDN, bandwidth is
not provided, rather it is a voice based data rate, 64 Kbps or multiple thereof. Consider this
contrast in terms of how cable TV companies provided voice communications in the early
1980's. Both Cox and Warner, using variations on LAN technology, delivered a voice, video, and
data service over the coaxial transport medium, by empowering the end users terminal, not by
regimenting the transport network.
Technology also plays a very pivotal role in telecommunications. Alfred Kahn (1971, p 300),
indicates that in the pre-divestiture period of the Bell System, the arguments for the needs of
both vertical integration and need for monopoly control were based on technology. Specifically,
there was a contention made by the Bell System that a single point of control to the network was
essential. Also, it was argued that an adequate scale economy was attained only through a single
monopoly. Indeed, given the state of technology of that time, the argument may have held. For in
point, the loaded copper transmission capabilities allowed only limited transport, namely one
voice channel per twisted pair. However, as we shall demonstrate, the underlying technology has
provided a dramatic change in the underlying system.
Functions now provided by the network, may be more efficiently provided by intelligent
Customer Premise Equipment (CPE). The question to be posed is; what is the role of the
network, and how do we provide the dimensions of creative freedom to allow these new roles to
evolve? To effectively approach this problem, we must first develop a canonical structure of a
network.
6.2.1 Elements
There are four architectural elements in the telecommunications network. These elements are the
control functions, the transport function, the interconnect function, and the interface function.
We now provide further detail on these functions. It should be noted that these functions have
evolved over the years in content and complexity. We view these elements in the context of a
communications network that must support the most advanced current concepts in
communications. Specifically, the world view adopted in this paper that lead to an interpretation
of this architecture are:
192 | P a g e
(i) End users desire to have interactions in a real time fashion with images and other high
resolution information that must be provided in a fashion that meet both time and resolution
requirements (See Barlow).
(ii) The end user devices are extremely intelligent and complex and can operate in a standalone
environment.
(iii) The users desire to operate in a totally distributed fashion. Data bases will be a different
locations, users are at different locations and input output devices are also at different locations
(See Dertouzos and Moses, and de Sola Pool pp 57-59 for details on these directions).
(iv) The network may provide different levels of service to different users. There is no need to
provide universal service of full capability to all end users.
This view of the network will significantly influence how extensively we defined the elements
and in turn will impact the combination of those elements in an overall architecture. All of these
assumptions on the world view are different than before, in an all voice world. In this paper, we
define a network as an embodiment of an architecture, in all of its elements.
The architectural elements are control, transport, interconnect and interface. The overall
architecture of the element interrelationship and the elements of the functions of the separate
elements. The details on each are described below:
Control: Control elements in an architecture provide for such functions as management, error
detection, restoral, billing, inventory management, and diagnostics. Currently, the voice network
provides these functions on a centralized basis, although in the last five years there have evolved
network management and control schemas and products that allow for the custom control and
management of their own network. Companies such as IBM, AT&T and NYNEX have
developed network management systems that move the control from the network to the customer
(McGarty, 87). On the sub-network side, companies such as NET, Timeplex, Novell, 3-COM
and other have done similar implementations for local area networks, data multiplexers and other
elements. Centralized network control is now longer necessary and in fact it may not be the most
efficient way to control the network.
What is important, however, is that network control providing the above functions is an essential
element for either a public or private network. Thus as we consider network evolution, this
element or set of function must be included.
Control has now been made to be flexible and movable. The control function is probably the
most critical in the changes that have been viewed in the context of an architecture. All buildings
193 | P a g e
need windows, for example, but where one places the windows and what one makes them of can
yield a mud adobe or the cathedral at Chartres. The same is true of the control element. In
existing networks, the control is centralized, but in newer networks, the control is distributed and
empowered to the end users. The users can now reconfigure, add, move, and change their
network configuration and capacity.
Let us briefly describe how the control function can now be distributed. Consider a large
corporate network consisting of computers, LANs, PBXs and smart multiplexers, as well as a
backbone fiber transport function. Each of these elements has its own control facility for
management and restoral. Each has the capability to reroute traffic from one location to another,
and the routing systems are programmed into the system as a whole. On top of these sub element
control functions is built another layer of control that views the network as a holistic entity. This
form of control has been termed a manager of managers.
It monitors all of the sub net elements and takes control if necessary. It is embodied in several
independent controllers, each having the capability of taking control from a remote network.
This form of organic network control has evolved in recent years and is now common in many
corporate networks. In addition, this concept of the organic network was described in detail by
Huber in the DOJ report to the U.S. Justice Department during the first Triennial Review of the
MFJ (See Huber).
Transport: The transport element is provided by the underling transport fabric, whether that be
twisted pair of copper, fiber optic cable, radio or other means. Transport should not be mixed or
confused with other elements of the network. Transport is merely the provision of physical
means to move information, in some form such as digital, from one point to another. At most it is
expressed in bits per second and at best it is expressed in bandwidth only. Bandwidth as a
transport construct is the most enabling. Transport does not encompass the need to change the
information or to do any other enhancement to the information.
In the early regulatory cases such as the Above 890 Decisions in the microwave systems that
were the precursors to MCI (See Kahn (II p12)), the Bell System argued that the technology of
transmission limited the transport to only those companies that had the transport, interconnect
and control. MCI on the other hand recognized that the customer was able and willing to
differentiate these elements of the architecture and would segment them in a more economically
efficient fashion. Specifically, in the early days of MCI, customers in the mid-west would select
multiple transport paths and would do the control function on their own premises. In addition,
the customers were willing to accept lower quality of service for a lower cost of service. The
lower quality was reflected by possibly a higher outage time.
194 | P a g e
It could then be recognized that the horizontal scale economies of all of the network elements,
including but not limited to transport, were actually diseconomies of scale in the market. (See
Fulhaber for a discussion of a more detailed view of scale diseconomies in terms of the new
architectural elements) Fragmentation and segmentation along architecture elements allowed for
the growth and efficiency of MCI. The emphasis should also be made on the statement of the
FCC Examiner in the MCI case who stated (Kahn II p 134), "MCI is a shoestring operation ... the
sites are small and the architecture of the huts is late Sears Roebuck toolshed." It is prescient to
note that the examiner used the term architecture for the microwave repeater sites when indeed
MCI was changing the architecture of the network. This remark is more than just an embodiment
of a metaphor.
In the current network environment, the issue of transport and its enabling capacity has again
arose. This has been the case with the introduction of fiber. Fiber may be segmented for the user
in terms of data rates or in terms of bandwidth. In the NREN, the three steps are all focused
along the lines of increasing data rates, from 1.5 Mbps to 45 Mbps to Gbps. As we have
discussed, bandwidth is the more enabling dimension, leaving the choice of data rate and data
structure to the end user. This capability is best deployed by using a dark fiber network. The top
network is a standard fiber network with repeater at periodic intervals. In current technology
limitations these are necessary because of the losses in fiber transport. However, with the current
state of the art technology, fiber can be strung for many tens of miles without such repeaters and
still maintain adequate transmission capacity.
Thus the repeaters are not there solely as a result of fiber constraints on transport. They are also
there because they enforce the voice regime of the voice based world view. Namely, the
repeaters do not repeat data rates, they also repeat framing sequences based on 64 Kbps voice
frames. Thus any work station must use 64 Kbps as the underlying data fabric. As an extreme
example, NREN in its Phase 2 will provide 45 Mbps to the users. Regrettably, there is no 45
Mbps modem. That is, direct access to 45 Mbps is not achievable. It must be sub multiplexed to
the equivalent of voice grade digital circuits. Thus the world view is pervasive in this design. The
same is true as SONET protocols are used in upgrades to broadband ISDN, especially over an
ATM switch (See Fleming for a discussion of broadband switching and the voice paradigm).
In contrast, dark fiber is the provisioning of an optical fiber to be used as the end user sees fit. It
is the world view analog of the LAN. The LAN provides co-axial bandwidth of several hundred
MHz whereas the fiber provides the bandwidth of GHz to TeraHz.
Interconnect: The interconnect element of the architecture describes how the different users are
connected to one another or to any of the resources connected to the network and is synonymous
with switching. Interconnection assumes that there is an addressing scheme, a management
195 | P a g e
scheme for the addresses, and a scheme to allow one user to address, locate and connect to any
other user.
Interconnection has in the past been provided by the Central Office switches. As we shall discuss
latter, this implementation of an architectural element was based on certain limitations of the
transport element.
With the change in the transport element of structures allowing greater bandwidth, the switching
needs have changed. Specifically, distributed systems and scale economies of the distributed
architectures allow for interconnectivity controlled by the CPE and not the Central Office. As we
shall show later, the advent of Local Area Networks and CATV voice communications are ones
using distributed interconnectivity elements.
Again, Alfred Kahn noted (II, p 127),
"We have already alluded to the technological explosion in communications after World War
II,...The case for a national telecommunications network monopoly has the following aspects.
Aggregate investment costs can be minimized.. if the planning for the installation and expansion
is done with an eye for the total system....Since any one of the 5 million billion possible
connections that the system must stand ready to make at any point in time may be performed
over a variety of routes....justifies the interconnection...completely dependent on its own
resources alone.”
This argument for interconnection, combined with transport and control (namely horizontal
integration) was valid in 1970. It however is not valid today. They are separable functions and
scale economies are in the hands of the CPE manufacturers not the network providers. In effect,
there exists no monopoly in interconnect as a result of these technology changes. This is a
dramatic change from 1971 and Kahn's analysis.
There are three general views of interconnection that are valid today; the Telcom, the Computer
Scientist, and the User. The Telcom view is based on the assumption of voice based transport
with universal service and the assumption of the inseparability of interconnect and control. The
Computer Scientist view is based upon the assumption that the network, as transport, is totally
unreliable, and that computer hardware and software must be used in extremis to handle each
data packet. Furthermore the Computer Scientist's view of the network is one where timeliness is
secondary to control. The Computer Scientists view has been epitomized in the quote, "Every
Packet is an Adventure". This is said with glee, in that each data packet is set out across the
network and it is through the best of hacking that the Computer Scientist saves the packet from
the perils of Scylla and Charybdis. The third view is that of the user, who is interested in
196 | P a g e
developing an interconnect capability that meets the needs and minimizes cost. This is
minimization of both obsolescence and cost strategy.
Processing cost or capacity is declining every year. Thus an investment must try to follow the
curve. In a hierarchical view of interconnect, such as a large centrally switched network, the
changes occur once every few years. Thus the lost cost or performance efficiency can become
significant. In contrast, in an end user controlled environment, with a fully distributed
architecture, the lost efficiency is minimized as technology advances.
Interface: The interfaces are the end users connection to the transport element. The interface
element provides for the conversion from the end user information stream and the information
streams that are used in the transport form of the network. For example, the telephone interface
for voice is the analog conversion device.
Interfaces were originally called "Alien Attachments". In Kahn (II p. 140-145,) he discusses the
history of the interface leading up to the Carterfone decision. The most significant position in
CPE control was the Hush-A-Phone debate from 1921 to 1946. The Bell System at that time took
total and full control over the quality of the delivery of the service of voice. The Hush-A-Phone
company provided a mechanical cup device that could be placed over the mouthpiece of the
telephone to assist in making the conversation more private. AT&T took the position that it
interfered with the network and the quality of service and battled this for 25 years. Such is not
the case today. CPE computer equipment has proliferated and the current costs for 9,600 bit per
second modems are comparable to high end voice telephone devices.
Clearly, this fourth architectural element is separate and apart.
We have divided the network elements into these four categories to demonstrate that there are
clearly four distinct and separable areas for growth and policy formation. Issues of regulation,
due to potential monopolist control are always a concern, but it will be demonstrated that in all
four there are economies in market disaggregation.
197 | P a g e
Network Providers
Type Customers Architecture
Public Switched RBOCs Hierarchical

Centrally Controlled
CATV Cable Companies Local Branched
Customer Local Campus Point to point

Point to Multipoint
Branched
Government Federal, State, Local Point to point
Point to Multipoint
Branched
Private Corporations Point to point
Point to Multipoint
Branched
McGarty, Harvard November 1990 Figure 2
Natural monopolies have been studied by many, and in the context of utility regulation there are
many key studies. In this context, Spulber has defined a natural monopoly as:
".. a property of productive technology, often in conjunction with market demand, such that a
single firm is able to serve the market at less cost than two or more firms.”
Natural monopoly is due to economy of scale and in the current architecture, elements of the
monopoly concept no longer applies. Potential monopolist tendencies in all of these elements,
separately, have been reduced by the ability of the end user to fabricate the elements of the
network in a set of separable fashions. Together these elements clearly demonstrate no
monopolistic power. The traditional theory on regulation has focused on the control over the
transport facilities. As we enhance the network structures and provide differing forms of
information transported, concern should also be focused on the other three elements (see Kahn p
II 127). In addition, monopoly power even over transport was based on the users inability to
individually justify the capital costs on the transport infrastructure. In certain cases, this is no
longer the case, finding many users easily justifying payback in less than one year on new
transport infrastructure
6.2.2 Architectural Alternatives
Is there a natural taxonomy for the set of network architecture alternatives? Do these present
limitations on what can be done or are they extensive? Is there a natural limitation in the existing
198 | P a g e
architectures that prevent the new technologies from introducing the new paradigms to the
communications world?
We address these issues in the context of several existing network hierarchies.
Hierarchical: The current network architectures are structured in a hierarchical fashion. As we

have already indicated, there are historical and technical reason for this architecture. We show in
199 | P a g e
Specifically, we see the set of transmission schemes connecting from a lower level to higher
ones. A path may or may not go horizontally. It may go vertically, all controlled by a single
control at the highest level.
Centralized: A centralized architecture is similar to a hierarchical system in that the control

function is centralized. However, the transport elements are not in a hierarchical format.
Centralized
INT TRAN SW TRAN SW TRAN INT
Control
CONTROL
Hierarchical
SW
TRAN TRAN
SW SW
TRAN TRAN TRAN TRAN
SW
TRAN
INT
200 | P a g e
The hierarchical structure is no longer present, but there is a single point of control. The control
element covers all other elements in the system. A typical example of this type of network is that
of a large bank in a metropolitan area. Part of the network is the local ATM (Automated Teller
Machine) network and the voice network for the bank. Each are separate but the bank controls
both from a single point of control.
Distributed: The distributed system has distributed control, distributed interconnection and flat
transport alternatives. Here we first note the reduction in concatenated switch and transmission
elements. The network is much less dense and the switch is actually co-located with the
interface. The LAN networks are typical example of distributed designs.
Distributed
INT INT INT
SW CON SW CONT SW CON
TRAN
SW
INT CON
Segmented: A segmented network is really a hybrid. Each segment uses a sub architectures that
meets the requirements of the existing system but the networks are interconnected through
standard interfaces.
201 | P a g e
Segmented
Central
INT
TRAN SW TRAN TRAN INT
SW
Gateway
Distributed
INT
SW TRAN SW
INT
In this case we show that this network architecture is an amalgam of the first three. What is still
common, however, is the partitioning into local and long distance nets. A typical example of this
network is that of a large corporate network. Part of the network can be for the voice circuits,
controlled at a single point and based upon use of both local and inter-exchange carrier circuits.
The second part of the network is the data network, again using both local and long distance
carriers, and control from a separate location.
Partitioned (Local and Long Distance combined in a community of interest): In all of the above,
we have assumed that local and long distance transport are separate. This is a world view
dominated by the regulatory environment. We can see the segmentation along community of
interest lines rather than along these more traditional lines. Thus one community of interest is a
network for financial service companies and a second for a network providing service to the
residential user. These each have all of the local and long distance services, but are now
segmented by the user market or the community of interest. The sub architecture may be any of
the above. This architecture allows for local and long distance in separate partitions. It says that
you can segment the network by users not just by function. Had the MFJ understood users rather
than functions, the results could have been dramatically different. An example of a Partitioned
network would be that for American Express or Sears. It contains the set of local and long
distance networks as well as subnets for specific distributed applications. However, each of these
companies may have access to a separate public switched environment.
202 | P a g e
Partitioned
Network 1
Long Distance
INT
SW
Local Local
TRAN Gateway
Network 2
INT
SW
Understanding that there are several varying architectural designs allows one to better understand
that each reflects not only connectivity but also the world view.
6.2.3 Impact of Technology on Architecture
We have just discussed the elements of the architecture and the embodiments of design that these
elements may lead to. We shall later discuss the details of the technology evolution but it is
appropriate at this stage to make several observations about the current impact of technology on
architecture.
In the current telephone system, the interconnect element of the architecture is provided by the
Central Office Switch and the physical interconnection of the wires from the street to that switch.
The point at which the many wires from the street meet the switch are at a device called the Main
Distribution Frame (MDF). The Frame must be able to connect any incoming wire to any
outgoing wire. The MDF, as it is called, has been the same for over fifty years. It is a manually
connected system , where the craft person must connect each incoming telephone wire to a
corresponding location on the switch, each time a customer moves or changes their phone
number. In computer systems, this is all done in an electronic fashion.
203 | P a g e
Infrastructure
Hardware
Network
Physical
Interfaces Virtual Logical Standards

Access Protocols
Relational
Data Formats
Addressing
In contrast, the central processing unit in computers goes through changes once every two years.
The standard processing capacity curves show a doubling of processing capability in the same
two year period. Computer users have a more rapid turnover of technology because they
generally work in an environment with no regulation, shorter depreciation schedules and a focus
on meeting specific business needs.
In contrast, the centrally based network must meet a collection of common needs and serve them
in a least common denominator basis. The conclusions from these observations is clear. If
change is at the heart of the services and technology is driving them, then migrating the elements
to the customer of control, interconnect and interface maximize the change and innovativeness of
the network.
204 | P a g e
Current Switched Network
Class 1
Class 2
Class 3
Class 4
Tandem
Class 5 CO Class 5 CO Class 5 CO
In terms of a national network, this then begs the question, should not the network, as
infrastructure, be nothing more than a broadband transport of open single mode fiber and let all
other functional elements be provided by the end user.
Consider what was written by a Bell System polemicist in 1977 at the 100th anniversary of the
Bell System at MIT. The author was John R. Pierce, Executive Director at Bell Labs, who
stated:
" Why shouldn't anyone connect any old thing to the telephone network? Careless
interconnection can have several bothersome consequences. Accidental connection of electric
power to telephone lines can certainly startle and might conceivable injure and kill telephone
maintenance men and can wreak havoc with telephone equipment. Milder problems include
electrically imbalanced telephone lines and dialing wrong and false numbers, which ties up
telephone equipment. An acute Soviet observer remarked: "In the United States, man is exploited
by man. With us it is just the other way around." Exploitation is a universal feature of society,
but universals have their particulars. The exploitation of the telephone service and companies is
little different from the exploitation of the mineral resources, gullible investors, or slaves." (de
Sola Pool Ed, Pierce, pp 192-194).
The readers should note that this was written nine years after the Carterfone decision and five
years before the announced divestiture. Pierce had a world view of an unsegmentable telephone
network. This paper has the view of a highly segmentable communications system. The world
view of the architecture has taken us from "slavery" of Pierce to the freedom of the distributed
205 | P a g e
computer networks of today. Kuhn has described technologists as Pierce as the "Old Guard",
defenders of the status quo. They defend the old paradigms and are generally in controlling
positions for long periods of time.
6.2.4 Architecture versus Infrastructure
It is important to distinguish between architecture and infrastructure. We have extensively

defined architecture in terms of its three parts; elements, world view and technology.
Infrastructure unfortunately has been reified in terms of some physical embodiment. The
discussion of NREN being an infrastructure is viewed by many as being a determinate thing.
Kahin has, however, de-reified the concept in terms of it being an embodiment of a concept or
set of common goals. We expand that and state that an infrastructure is an enabling capability
built around a common construct.
Fiber Distributed Network
Dark Fiber
E/O
E/O Tera Bit Maximum
There are four types of infrastructure views that are pertinent to the current discussions of
networks. These are of particular import to such networks as NREN since they will lead to the
policy directions that it will take. These four infrastructure types are as follows:
Physical: This is the most simplistic view of an infrastructure. It requires a single investment in a
single physical embodiment. The old Bell System was such an infrastructure. The National
Highway system is such an infrastructure.
206 | P a g e
Logical: This network may have separate physical embodiments, but all users share a common
set of standards, protocols and other shared commonalities. All users have access through an
accepted standard interfaces and common higher level transport facility. IBM had attempted in
their development of SNA in the mid 1970's to develop a logical infrastructure in data
communications. This was expanded upon by the ISO OSI seven layer architecture, selecting a
specific set of protocols in each layer.
Virtual: This type of infrastructure is built on intermediaries and agreements. It provides shared
common access and support interfaces that allow underlying physical networks to interconnect to
one another. Separately, the individual networks may use differing protocols and there are no
common standards. The standards are at best reflected in the gateways to the interconnection of
the network. Thus this infrastructure is a loose binding through gateways. It is in many ways
what is the INTERNET today, if we include all of the subnets.
Relational: This type is built on relationships between the network parties and the establishment
on higher level accessing and admission. Specifically, a relation infrastructure is based on
agreements on sharing addresses, not necessarily common addressing, and on the willingness to
share data formats and types. It is an infrastructure based on shared common interests but not
shared common access. This type of infrastructure is what in essence exists in most cases today.
Users can move from network to network through various gateways. The difficulty is the fact
that the interfaces are cumbersome and may requires sophistication on the part of the users.
However, more intelligent end user terminals and interfaces will reduce this cumbersome
interface problem.
We show the relationship of these four infrastructures in a diagrammatic fashion. Our conclusion
is that understanding the type of infrastructure that the coalition of users want, will also impact
the architecture, based upon an imputed world view. Arguably, a physical infrastructure leads to
maximum hierarchical control and the resulting impacts that such control leads to. This is a
critical issue for networks such as NREN, since by choosing infrastructure and architecture may
not be as uncoupled as desired. In particular, the selection of Gbps capability may really be GHz
capability and is best suited to a Virtual or Relational infrastructure.
6.3 TECHNOLOGICAL FACTORS
In the previous discussions, we have assumed that there are certain underlying stabilities in the
transport structures that enable the separate network providers to perform their tasks. There is a
growth in technological capabilities that may cause dramatic changes in structures that we have
discussed in the previous sections. In this section, we will focus on some of the dramatic
technical changes and discuss the impact that they may have on the market equilibria established.
207 | P a g e
6.3.1 Transport Capabilities
Transport is the raw power to move information from one place to another. Transport is also
viewed in its most primitive form, specifically bandwidth, rather than data rate. In current
systems, transport has significant capital in the twisted pair plant as well as the fiber backbone.
There are several alternatives to twisted pair that are evolving and we shall discuss their
directions briefly. These directions may significantly change the view of capital allocation to the
transport portion of the network.
Specifically the raw transport capabilities have been dominated by the capital cost of a twisted
pair, with that single pair limited to a single voice channel. There are at least three technology
areas of change that may impact the capital asset allocation equation. These areas are
developments in fiber, radio and surprisingly in twisted pair itself. One of the goals of transport
development is to have "Free" bandwidth access, or as close to it as possible, in relation to the
other three network elements.
Under the current structures, this bandwidth is so segmented that it is impossible to foresee a free
state.
Another factor in increasing the capital costs of the existing network is the need to add capital in
large amounts and not in incremental amounts. Thus, a central office is added with the capability
to handle several hundred thousand users. In contrast, with a fully distributed network, it is
possible to add user capital as each user is added. We have discussed this issue in the last section
and demonstrated that in an environment of rapidly changing technology, the cost-performance
curve versus time is rapidly improving, so that small incremental changes allowed by fully
distributed system optimize economic performance.
The following are three technological factors that will lead to the goal of freer bandwidth.
Fiber; Uses, Users, and Costs
Fiber has revolutionized the data networks in the United States. A single strand of fiber can
transmit 10 12 bits per second of data. If we allocate each home, 100 million residences, with
100 Kbps of full time data, that is 10 13 bits per second if everyone in the US is talking
simultaneously in this high speed data fashion. That is the capacity of just a single strand of
fiber. A typical bundle of fiber has 25 to 50 strands and these are connected to other such
bundles. The current fiber network is structured like past voice networks, and generally does not
take advantage of the bandwidth of the fiber. Albeit the technology is not yet totally
operationally capable, the world view of the system designers is one that is to use fiber as
copper. Use it for one voice circuit after another.
208 | P a g e
As we have discussed, dark fiber is the most flexible form of access for the more sophisticated
user. It enable the user to maximize the impact of processing power and it optimizes the tracking
of the cost performance curves. However, it is a fully distributed system and thus requires the
careful control of the infrastructure relationships established. It should also be noted, that
although the fiber loss characteristics necessary to support a wide dissemination of dark fiber are
theoretically available, this is only in a laboratory setting. There are no commercially available
fiber systems to allow this in today's network.
Radio Spectra; Access, Capacity and Cost
The current cellular system is being supplanted by new digital cellular technology that can
support in excess of 100 times the number of voice channels in the same spectrum. It will be able
to support data rates of from 1 Mbps to 100 Mbps. There is pressure to allow the cellular
companies to widely deploy this technology and it has been estimated that the capital costs per
cellular phone will become less than the capital costs of a wireline phone. The change in
technology may make the replacement or build decision not one between fiber or copper, but
between copper or cellular. The policy issue is the effective use of bandwidth. If the FCC, in the
Common Carrier Bureau, and the FCC, in the Radio Carrier Bureau, can arrive at two different
decisions, then there will clearly be a significant policy debate.
The two positions are as follows. First, assuming that the radio bandwidth is free, then if the
new cellular technology is less capital and operational cost intensive then copper or fiber, should
the FCC allocate the bandwidth in the public interest. If the answer to this is yes, then how
should the FCC allocate this resource in the most competitive fashion. Does the current Cellular
policy of two operators per system still hold, or should there be an imputed cost to the bandwidth
in a bid process. If there is an imputed cost, who is to receive the economic benefit.
The radio technology will allow many others to enter the market, restrained only by the
bandwidth limitations on available spectra. This represents a potential destabilizing technology,
especially for residential and rural networks.
Twisted Pair; Utilization, Capability and Sunk Cost
As we mentioned before, twisted pair has been limited by culture to single voice channels or 64
Kbps at a maximum rate for data. Current advances allow for transmission at rates of 100Mps on
unshielded twisted pair. Thus it is possible to continue to utilize much of the existing copper
plant for the types of data services that are required by many customers. Therefore twisted pair
should not be considered defunct.
6.3.2 Interconnection
209 | P a g e
Interconnection is the architectural element that provides for the function of allowing each user
to interface with other that are connected to the network. In the existing telephone network,
switching as we know it was introduced since bandwidth was very expensive. It was not
primarily introduced for the purpose of interconnection. Thus the central offices of today are not
there for the sole purpose of connecting one user to another. They are there, primarily, for the
purpose of concentration on trunk circuits, and thus preserving bandwidth. If one looks at the
development of the communications technology through the Electronic Switching Systems, the
intention was always to utilize the voice channel trunks at as great a level as possible. It was not
just to interconnect one user to another.
Past interconnect technology began first recognizing that bandwidth was expensive and that
concentration on the trunks was also essential. In the past, a twisted pair could support only one
voice channel. A central office could readily connect one local user to another local user through
the switch by merely closing a crossbar relay or a fereed switch. For connections to other central
offices, trunk networks were needed.
Since copper was expensive but of very limited carrying capacity, it was necessary to design the
switching or interconnection network to first minimize the need for copper trunks and then to
assure that full interconnection could be achieved. The view of the network as a hierarchical
design was a world view based upon the paradigm or technology of copper twisted pair, of
copper as a limited bandwidth transport vehicle.
Changes in technology show that interconnection can be migrated to the customer premise. As
we have indicated, the capability of a single fiber is adequate to handle all of the telephone users
in the United States. The interconnection in this system may be done by assigning each user a
separate frequency and using a laser tuned circuit to perform the switching function. Thus the
switch is at each customer telephone, and all that is necessary for the transmission function is a
single strand of fiber. Point of fact, there is research under way at MIT and other institutions to
develop just such systems. In extremis, this approach reduces the public switched network to a
commodity based transport only facility. If we accept the validity of this alternative world view,
then we can dramatically see the changes that may occur in the national network.
Central Switched Networks; Monopolistic Necessity
The central switched network was the result of the bandwidth preserving approach of the
nineteenth century. It was further reinforced by the regulatory emphasis on rate of return
regulation. In addition, this approach assured a barrier to entry to any form of competition.
210 | P a g e
The architecture of the current centrally switched networks starts with the Class 5 central office
switches, moves up to Class 4 trunk or tandem switches and migrates even up to a class 1 switch,
which handle only excessive overflow traffic.
Fully Distributed CPE Based Interconnect; Free Bandwidth
The first entry into a distributed interconnection capability was the local area network
technology, LANs, particularly Ethernet connections. The Ethernet connection provides a fully
distributed interconnect capability. In the Ethernet configuration, each user has an Ethernet card
in their machines and it is in this card that the signaling on the Ethernet channel carries the
signaling and interconnection information. The signals have information of the source and
destination of the packet message. This is read and decode by each of the terminals on the
network.
Ethernet was the first architecture to challenge the existing hierarchical architecture in the
interconnect element. It empowered the end user to control the interconnect function directly and
enhance it as the needs required. With the introduction of routers, bridges, and gateways, more
complex inter-network connections can now be achieved. Thus LAN intensive companies can
inter-network all of their locations on a fully switched basis by use of CPE based systems, and
use only the direct point to point transport of the common carrier.
As we see the expansion of the bandwidth on a fiber loop, the capital investment of fiber per unit
data is de minimus. For example, if the fiber costs are $50,000 per mile, and one can carry even
one Terabit per second, or 10 12 bits, then the cost per bit per mile is $5 10 -8 per mile. That is it
is one the order of a millionth of a cent per mile. If we use a million miles and the capital costs
are a cent per bit per second. In effect, technology is driving the bandwidth costs to zero, as
compared to other costs, We depict a simplistic view of a fiber network. In this network, we
envision that all of the transport is in the network and that the interconnect, control, and interface
are part of the CPE. This is a Terabit per second network that allows the end user the maximum
in interface capability. The end user may access any and all of the data capabilities of the
transport network, not limited by switching or other elements.
This technological change will undoubtedly change the world view of network providers. If
bandwidth is relatively free then the use of processing at the CPE becomes a vital ingredient in
the network design. The hierarchical network is no longer the only choice, in fact its viability is
called into question. The move towards that end, as we have already shown in the LAN area, is
already under way. The development of new means of interconnect will result in significant
changes to the network. Specifically, we see;
211 | P a g e
The network will become more organic. The end user will have direct control and access to the
interconnect, interface and control functions.
There will be less materiality of scale. In the current architectures, there is scale and the
performance of interconnect is a result of the scale of the network. Simply stated, we need a big
network with a great deal of switches so that we can talk to one another. In the network with user
controlled interconnect and "free”
bandwidth, the materiality of this scale is no longer a factor. That is, infrastructure is at best
irrelevant and at most counterproductive.
Multiple overlay networks are connectable, from both within and without the core net. Thus,
viable overlays can lead to local short term optimizations that meet end user needs.
o Intelligence in the CPE is expansive and reduces the capitalization needs for networks. It also
reduces the time scale factor for the introduction of new technologies mapped to the technology
change curves.
6.3.3 End User Interfaces
The use by the end user is the driver for new network applications. That end user must have
some economic justification for the use for it to be meaningful. It must fit the value chain of the
user or of the customers. Unless this is the case the result is a technology in search of a market.
There are four elements of the end user interface that must be explored, the first the physical
constraints and the later three the application specific elements. Specifically, the user interface
must first be a functional input/output device, displaying information in its broadest sense as well
as enabling the user to interact with all elements connected to the network.
For the most part, the last ten years has seen the evolution of the end user interface from a dumb
terminal to an intelligent, highly interactive, display device. It will be in the end user interfaces
that the new drivers for network usage will arise. As we have indicated before, the world view of
the telecommunication network was based on a voice modality of instantaneous
communications. The new modalities are focused on multimedia and multi user communications
systems. There will be an integration of video, voice, text, image and other sensory interfaces.
More importantly, there will be more processing power in the terminal to perform the tasks that
the network now performs best in its hierarchical way.
Terminals and Displays; Uses, Access and Price
The evolution of the end user terminal has been and will continue to be the major driver in
changing the architecture of the network. The use that is made of the conversational nature of
212 | P a g e
communications, the sources and structure of the dialogue, will combine multiple media
modalities. The second trend is that of a more intelligent end user terminal. This intelligence is
not only in the physical processors in the terminal but also in the distributed software capabilities
that reside on top of the device. A fully distributed operating system environment, as well as a
fully distributed database environment, will allow for highly interactive communications
environments. Users will be able to share information of significant data content in a fully
distributed and real time fashion.
Moreover, the functions of network management, network administration, and interconnectivity

will be controllable at the terminal end and not reliant upon a single hierarchical network. In fact,
the structure of a regimented network with its hierarchical structure may actually inhibit certain
types of applications development. For example, in the areas of multimedia image
communications, the need is for as great a bandwidth as possible, with as little network control
as possible. The use of the OSI lower layers, such as data link, network and transport, actually
add significant delays into the communications link. For example, in the transfer of a compound
multimedia image of a 100 Mbit image, a set of four synchronized voice signals, and a video
segment, the use of either an ATM switch or SMDS switch, using cell transfer rather than full
frame transfer, can result in delays exceeding five seconds. The packetization required seriously
effects the overall response time and performance.
Information; Target Markets and Economic Returns
Information networks have evolved over the past fifteen years with the early introduction of the
database services of such market leaders as Lockheed, with the Dialog system. This industry has
evolved along market niches, providing such markets as the financial services market with a
wealth of information sources. Companies such as Reuters, Knight Ridder, Quotron, and others
have made this segment a billion dollar industry. However, there has been little or any interest in
the consumer side of the information market. Information is a strategic element in a corporation's
competitive advantage in the market. It is not, and has not been a major element in the
consumer market. The consumer is focused on entertainment and not information.
However, the evolution of the network architectures will depend upon and will influence the
information interfaces to the end user. Currently, there are significant delays on the Dialog link
due to limited transport and the single location of the database storage. With a fully distributed
architecture, the Dialog system can be moved from its current location on the San Andreas Fault
and become distributed in multiple locations. Users will be able to obtain real time full text
information for off line perusal.
213 | P a g e
Major Technology/Service Drivers
Transaction
Choice Entertainment Promotional

Information
Selection Persuasive
“TIES:
Electronic Distribution Channel
Information can also be provided in a distributed form, and no longer in a hierarchical form.
Transactions; The Electronic Distribution Channel
Judge Green has looked to Gateways as a means to enter the information age. As noted,
information services are naturally a valued added benefit to commercial users and the consumer
is generally interested in entertainment or at most a limited transaction capacity. Transactions are
also purchases and to obtain the purchase the seller must not only have access to the buyer but
also must promote and persuade. The video medium that currently exists is a promotional
medium and creates awareness. It is still questionable whether it is persuasive.
The vision of Videotex has been significantly blemished over the past ten years in the United
States. We all too frequently look to the success of Minitel in France, but we all too often have
no knowledge of the French Telephone System. Telephone Information Operators in France are
almost nonexistent and when available have been known to hang up on customers in less than the
most polite fashion. In addition, the French Telephone company, in conjunction with the French
Newspapers, developed a way to ensure a noncompetitive advertising market in an electronic
fashion. The classic yellow pages were abandoned, the terminals were underwritten by the
government as a social action move to develop a technology base, and the public turned to their
favorite pastime, now in digital form.
214 | P a g e
Minitel is not the harbinger of things to come in the area of home information systems. The
regional Phone companies have all introduced videotex gateways for use of information services.
These have been of little success. It is still to be determined what the consumer wants from
electronic information services.
As indicated, the use of the communications channel through the end user interface allows for a
distribution or sales channel for products or services. Promotion and persuasion are critical and
these require high quality, personalized, self-segmented, full motion video. The impacts on the
network are significant and it is possible to perform this task only with the extended network that
we have developed.
The architecture of user empowerment is essential and existing networks cannot meet the needs.
Entertainment: Generator of Revenue
The entertainment interface has been the television set and the transport mechanism has been the
coaxial cable of the CATV operator. Fiber to the home or other possibilities are dependent upon
the entertainment driver. As we shall discuss latter, the CATV provider has a secure monopoly
of service. The entertainment interface will be driven by that supplier.
Totality of Interfaces
We have dedicated the totality of interface elements. It is a conjunction of transaction,

information and entertainment, overlaid on displays and input/output devices. The focus on the
residential side will require an amalgam of these factors. The convergence on the commercial
side will require transactions and information. The transaction element is key to the overall
success. Thus in evolving networks, we should not disregard the transaction portion.
The key conclusion from this section is that technology has changed and with the power to
perform most if not all of the network functions placed in the user's hands, at drastically different
cost factors, the need for infrastructure, viewed as a single logical reality, is no longer necessary
or possible. In the next section, we shall show the need, first, to meet the market demands, and
second to create value for all users.
6.4 MARKET ENVIRONMENT
The primary driver for the development, application and usage of any resource or asset is the set
of market elements who view this asset as a means to increasing value in some form. We assume
that there exists a rational consumer and that in turn the consumer has a set of utility functions
that can be maximized or optimized by use of this exogenous asset (See Henderson and Quandt).
Understanding the current segmentation of the market by use and user allows for a rational
215 | P a g e
approach to the focusing of additional assets by the provider of the communications
infrastructure (Porter, 1980, 1985). However, one must be careful of the possibility of a circular
reasoning that says that the current market is the basis set for extending the future market. The
current market is conditioned by the existing trends in prices, accessibility, and performance. If
this is changed, we must be careful to assess the change in the future market. Specifically,
Mandelbaum has indicated that the users become de facto market makers and that they are
empowered by the unbundling of the network elements. The market maker influence is critical in
understanding the dynamics of the network architecture revolution.
Needs/Benefit Cycle
Supplier Needs Change Distribution

User Benefits
Increased Efficiencies
Convenience
Improved Distribution
Expanded Choice
Protect Market
Promote
Maintain And
Growth Educate
Supplier Benefits User Needs

Increased Business More Goods
Expended Distribution Increased Access
Protected Market Increased Demand
Value creation and transfer is the key construct in understanding the market and its relationship
to the new paradigms created by broadband networks. Value creation and transfer (See the
discussions in Fruhan in terms of the quantitative measures of value for users). Value can be
measured if the impact of the new technology or paradigm is understood in terms of its impact
upon the user of the new technology. Meeting the needs of the end user is creating this value or
the transfer of this value. Understanding the market is truly understanding the needs of the user
as the user perceives them and not assuming in a Kantian sense what the user "should need".
216 | P a g e
User Value Chain
PROVIDER USER CUSTOMER
Revenue Revenue Revenue
Profit Profit Profit
Admin
R&D
SW
Dev Ops
Profit=Revenue-Revenue Driver*Productivity Factor*Unit Cost
In the current market understandings of the market even by regulators there is a change in the
understanding that value to users must be created. For example, Noam has indicated that:
" Perhaps the greatest common failing of these traditional organizing ways of looking at the
telecommunications principles is that they concentrate "supply-side" analysis. That is, they look
at the subject from the angle of the production and the producers... Thus, one should not view
deregulation as a policy of primarily liberalizing the entry of suppliers. Just as importantly,
though much less obviously, it is the liberalization of exit, by some partners, from a previously
existing "sharing coalition" of users which has become confining.”
Clearly, the need for user empowerment, the formation of new coalitions and the establishment
of infrastructures unbound by the physical only are critical. It is essential to look at
telecommunications from the users standpoint and not the suppliers. Deregulation is not, as
Noam points out, the freeing of the suppliers as much as it should be an empowering of the users.
Thus, in this section we discuss the dimensions of the market makers and the market players. We
also note that regulation all too often controls the providers and all too often dis-empowers the
user.
6.4.1 Market Players
The market players represent the segmentation of the users of the telecommunications markets
assets.
217 | P a g e
They are motivated to do so in a rational fashion and do so as to maximize their utilization
criteria. For example, if the set of market players are large corporations, then the use of
telecommunications is that of a strategic asset to increase their overall competitive advantage.
That advantage may, and often is reduced to a simple set of financial returns. In contrast, the
consumer or residential user tends to maximize a utility function that is much less well defined
and is of increased complexity. It relates to alternative expenditures for such items as leisure and
entertainment.
Commercial User
The commercial user of communications network generally is a corporation who is using this
communications capability for either strategic or tactical purposes in their business. Such
companies as American Airlines have used the communications networks as a means of
capturing market share through the concept of the electronic marketing and distribution channel
(See Konsynski and McFarlan for a discussion of how this may expand beyond just a single user
into a coalition of users. This builds on the concept of Noam as well as that of de Sola Pool). The
tactical application is the use by companies in the operations of point of sale services.
The commercial user's market dynamics are most easily understood. Value creation can be
measured because the commercial entity has carefully analyzed and studied the structure of the
business and can immediately measure the impact of new means of productivity.
Consumer User
The consumer user of communications has the lowest level of expectations in terms of the
delivery of services. This user is expecting at best adequate local and long distance
communications. The suppliers of communications view this segment as a significant source of
new revenues. The CATV companies use their access channels as means for entertainment
distribution, only (See McGarty & McGarty, 1982, for a discussion of the market needs for
CATV). They offer a level of service and a level of expectations that is dramatically different
than that of the telephone provider.
Consumers have a complex set of benefits. This set is often not readily measurable and there
may not be a ready realization of the quantitative benefit. The concept of a utility function or
preference function defined for the consumer on an individual is generally quite difficult to
measure. In contrast, in consumer markets, market segmentation allows for effective and
efficient measures of impact. This will allow for a measurement of the impact of such new
products and services that broadband can bring to the consumer.
218 | P a g e
However, past experience, as we have already discussed, has shown that consumers are quite
complex and their behavior in the use of electronic elements has been difficult to predict and
variable in its response.
Corporations who have based their future on understanding the consumer have failed in repeated
cases in developing electronic services to the consumer. This is a clear warning to academics
who attempt to think that their needs are reflective of the masses that make up the consumer
market. One can simply imagine an MIT student designing a consumer product, assuming that a
UNIX (TM, AT&T) interface and use of LISP is essential because it gives him/her the best
response.
Government User
The government user is focused on meeting policy driven directives. The government user is
generally slower to act, acts in a highly competitive cost efficient fashion, and is generally risk
averse. In addition, the government user must deal with large coalitions of users with often
conflicting goals and objectives.
This market represents the most difficult to deal with as a result of the internal stress.
6.4.2 Market Drivers
The market for communications services is driven by the needs of the end users. These clearly
are different for the three market segments that we have just described. market drivers are those
factors that lead users to demand and use services. The market cycle is shown depicts the need
benefit cycle for the user and the provider. This cycle is the same for both the commercial and
the consumer user.
For example the cycle has the following dynamics (See McGarty, 1989):
First, The supplier has the need for new revenue. This need is recognized and an investment in a
changing distribution capability is created.
Second, the user is provided a certain set of benefits. The user is first to benefit. In some cases,
even though the supplier of the service desires to benefit, he must first provide that benefit to the
customer.
Third, these benefits convert into user needs. The user takes time to recognize the benefits and
then to internalize them into needs. Needs the create demand. A need provides the basis for a
sustainable demand.
219 | P a g e
Fourth, the user needs convert into supplier benefits. This occurs latter in the cycle. It is critical
to note the significant time delays in this process.
This cycle takes time for new and innovative products or services. This cycle may also be broken
at any single point and result in the dissolution of the specific market opportunity. However, the
basic concept of the need-benefit cycle still holds true. The dynamics of the cycle impact the
value of innovations to all players. Delays and the impact on imputed interest costs will increase
the need for payback to the investor.
Thus the cycle and its dynamics show that value creation results in lower value as the time to
complete the cycle increases.
Consider the example of the development of electronic banking, specifically the use of ATMs
(See McGarty, 1980, 1981, for a detailed discussion of this technology and its early
development). Initially banks wanted or needed to have less costly means of service their
customers. In addition, the banks viewed this electronic marketing channel as a means of
attaining greater market share based on the concept of location. This service was not needed by
the consumer at first but it did provide a set of benefits. These benefits were targeted at the select
share of the market, namely the young and well to do segment. By having the ATM network, the
bank could provide this segment of their market with the benefits of cash at any time and at any
location. This consumer benefit then changed into a consumer need and the broader base of
banking consumers requested the service. Then, this resulted in a direct benefit to the banks as
the size of the electronic banking led to them (See Zuboff, pp 132-145). The same will apply to
accruing the benefits from use of the NREN. The market drivers then are the factors that meet
the needs of the supplier and the benefits of the user.
We can expand this into the needs for broadband communications services. As we discussed
before, NRI is interested in developing broadband, as defined by Giga bit per second data rates,
from a national infrastructure perspective. Their initial focus is on national supercomputing
networking. In contrast, there are many applications for imaging that are directly related to
solving end user requirements of current interests. For example, the printing, publishing and
advertising industry is going through an major infrastructure change. There are Apple Macintosh
computers in use for ad composition, editing, and layout. There are digital prepress systems that
allow for digital printing. There are not, however, the electronic means to connect these two
systems. If there was an effective system, then advertising agencies could cut operating costs an
estimated 5%, and press time could be shortened several hours. These are significant impact
factors in this business (See McGarty, Nov., 1990).
A second current application is in meeting the needs of the medical imaging community. For
example, if a large hospital can take the responsibility of reading the x-rays for a local HMO, the
220 | P a g e
current costs are on the order of $60 per patient per procedure. It has been shown that just
clustering the readings have a scale economy that reduces this cost to $40. If we further include
and electronic imaging system with a Hospital Information system, these costs drop to $20 (See
McGarty and Sununu, 1991, for a detailed analysis of the impact of the use of imaging
technology in a Health Care environment. The authors detail the cost savings directly attributable
to this technology and presents results from actual operations. The authors have also developed
an extensible methodology for determining cost benefits in this type of broadband environment).
These two simple examples show that there are clear and present opportunities for broadband
that are local and that, if properly structured, meet the market needs. These two examples also
result in two conclusions. In addition, these examples show that customers can and do, today,
create their own infrastructures using the new technologies. In the small, this clearly shows that a
global infrastructure is not only not needed but would be redundant. Specifically:
End users drive technology through their value chain productivity gains. Lowered costs or better
competitive positioning are essential. Broadband provided scale economies to businesses that
heretofore may have had little.
The development of broadband does not require a large infrastructure. It can be built around
focused local networks and applications. The needs of the users are not just transport. The
interfaces and interconnections are more significant drivers in an information rich environment
such as imaging.
Moreover, these examples have shown that today's transport, control, interface and interconnect
technologies meet current needs. The migration from current systems to broadband is possible
and achievable and are also economically efficient. Speculative applications based upon non-
market drive idealizations will clearly not replace currently customer directed and driven
applications.
6.4.3 User Value Chain
The generation of value by a user has been discussed in a dynamic sense as the creation of value
by increasing productivity on the part of a user or allowing for the development of new revenue
sources.
Value was defined in terms of the increase of the flow of funds to a firm by performing a specific
task.
The value chain concept, in contrast, is a static view of value projected back onto the operational
elements of the firm. It is a key concept in the full grasping of the value flow to a firm with the
introduction of a new technology. The value chain analysis, as developed by Porter (Porter,
221 | P a g e
1980, 1985, 1990), is a construct that overlays the rational utility function maximization process
of the commercial user. As we indicated in the preceding paragraphs, we will focus on the
commercial user because their utility function is generally more evaluateable and can be readily
related to a rational decision process.
We show the provider, the user and the customer. This is the natural food chain of the economic
market place. The provider must provide the user with effective supplies necessary for the
production and delivery of goods. The user, to attract a customer, must also provide the customer
a similar set of benefits. If we view the revenue of the user as the size of the total box, and the
expenses as their corresponding areas, then the users profit is the area left over after all expenses
are taken care of. The revenue is provided by the customer, the expenses controlled in part by the
provider. The company, namely the user, spends money on horizontal elements such as
Administrative functions, and vertical elements such as Software development. The allocated
profit is represented by revenue less expenses in each segment. The expenses are a product of a
factor driven by the customer demand, the revenue factor, the company's productivity, and a unit
cost. Thus for a fixed revenue, profit is increased by lowering costs or increasing productivity.
The information networks that we have described are productivity enhancers, thus profit
enhancers. This is the essence of the Porter theory.
The value chain concept views the user as an operating entity with sequential and simultaneous
operations as part of running the business. The sequential operations follow the flow of goods
into the establishment, through the processing done to add value and then out of the
establishment. The simultaneous operations are followed over all tasks and may include such
functions as finance, legal, and marketing. The company can then allocate costs and value to
each of these elements, and then can compare them to its competitors.
The costs of each step in the process are the result of three factors; the revenue drivers, the unit
productivity, and the unit costs. The use of information or communication networks allow the
user to improve the productivity or reduce the costs. This allow for increased competitiveness
and thus better margins. If the seller recognizes the value chain of the buyer, then the product
that is sold can be positioned in a similar fashion, thus helping the buyer to improve their value
chain. This will increase the revenue to the seller.
The value chain analysis provides a methodology to integrate the effects of communications and
information services into the evaluation of a business. Porter has done this for many segments of
many industries and McGarty (1989) has developed a detailed micro model to use in a detailed
competitive analysis. As we look at the market factors, value chain theory states that the use of
any new technology must be evaluated in terms of not only the end users value chain but also the
value chain of their customers. The chains are linked and the effect is complex
222 | P a g e
6.5 GOVERNMENT NETWORKS
Recent changes have seen a migration of Government networks from the totally owned and
operated FTS to the leased FTS 2000 evolve. It will be argued that the Government, as a
customer, is neither a key player in defining architectures nor do they drive the trends in new and
innovative systems and services.
In fact, as we shall demonstrate, they are at best the preservers of the existing world view in an
attempt to minimize risk and reduce perceived increased costs.
However, before continuing, there are several factors about the Government nets and their past
achievements that are appropriate. Dr. Robert Kahn, formerly Deputy Director of DARPA,
husbanded the ARPA net through its infancy (See Markoff, 1990). During that phase, he
approached AT&T to work with DARPA on the new and emerging packet technology. Kahn
indicated that AT&T not only failed to understand the technology at that time but further were
very slow in providing the Government with the information to use the existing modems that it
provided in its existing network. Thus Kahn, despite the position of AT&T, developed a true
infrastructure network. Out of this evolved the current packet nets of today. However, these
networks have been supplanted in many ways with other technologies, primarily those on the
customer premise, such as LANs, combined with user specific virtual networks. The reason for
the change was that the original premise of the packet networks usefulness as an infrastructure
which was that of cost reduction for the computer user. With the breakup of AT&T that equation
changed and now users have a variety of cost performance tradeoffs that are both available and
more attractive.
Currently Kahn is moving in other directions, as head of NRI (The Corporation on National
Research Initiatives), a not for profit research consortium, he is pursuing the vision of a
broadband network, called Aurora, that will build on a research consortium of various
universities. The goal is to interconnect supercomputer over a Giga Bits per second network. In
the ARPA Net days, the goal was clear; computer to computer communications in a cost
effective fashion. In the current scenario, the goal is more elusive.
However, the world has changed. The world now is one of high competition, intelligent
terminals, and a migration of many of the networks old functions into the end users devices. Can
even such a government infrastructure be justified in light on the underlying equation? In
addition, the drivers of networks today are the applications that the end users are to put on the
network, not just needed to transfer large computer data files. Thus one is faced with the
quandary of placing a sophisticated network in place to challenge the applications developers, or
to let the applications developers stress the network to its limits using the driver of customer or
end user demand. A research network must be flexible to stress the technology but not too costly
223 | P a g e
to result in preordained failure. The major accomplishments of the ARPA Net were electronic
mail and layered communications protocols.
The former was what every Net user found of use on the net and the layered architecture was a
natural offshoot of all of the interfaces required to meet some form of flexibility.
6.5.1 Structure
The Government networks can be divided into federal and state networks. For the most part we
shall focus on the Federal networks. There are two extremes in these networks, DoD based
networks and non DoD efforts. We shall not focus on DoD because of their special infrastructure
requirements. Non DoD networks have been focused on the meeting of a wide set of standards.
They must balance the needs of the voice user and the data user. Further, there is currently a need
for both interconnection as well as ensuring security in the network. For the most part, the
Government networks have outsourced the management and control. The control functions, such
as billing, are eliminated by the bulk buying characteristics.
6.5.2 Competitive Environment
Clearly, the Government networks are in a noncompetitive environment. At best they are limited
market makers as described by Mandelbaum
6.5.3 Optimization Criteria
Performance in the Federal Networks is best exemplified by the FTS 2000 effort. The
requirements for the network were developed and the selection of the provider based on cost
competitiveness.
6.5.4 Evolutionary Constraints
The major evolutionary directions for the Government networks are based on meeting their user
coalitions needs while doing so at a minimal cost. In certain cases, such as in DoD networks,
advanced technology will be employed. Other than this specific case, they will generally be
followers of other network leaders.
6.6 PUBLIC SWITCHED NETWORK
The Public Switched Network is the result of an evolutionary process that had given rise to the
Bell System and has resulted in the current structures under the Modified Final Judgment (MFJ)
strictures (Coll; Geller; Kraus and Duerig). The network is based upon a set of technologies that
require massive investment in capital assets to allow for the interconnection of many users to
224 | P a g e
each other. The system requires massive switching infrastructures since the underlying transport
facilities, the classic copper cables called "twisted pairs", have very limited information carrying
capacity.
6.6.1 Structure
The public switched network is built around a hierarchical architecture that makes several
assumptions of its environment and the customer base. These assumptions are (See von Auw for
the insider Bell System view, pp 334-398; see Toffler for the external consultants vision of
where the Bell System could have gone):
Bandwidth is a costly commodity so that it is necessary to provide for concentration of circuits

on trunks and tandem lines. This concentration leads to the need for multi layered switching
centers, a Class 5 Central Office being the lowest level.
Voice is the primary means of communications and all circuits are to be considered multiples of
voice circuits. In addition the voice is to be sampled at a rate of 8,000 samples per second and
the number of bits per sample may be from 7 to 8.
Universal service is necessary in order to meet the needs of the state regulatory bodies, This
means that telephone service must be structured to cover the gamut of the rural home to the large
corporation.
Quality of service is to be as high as possible with overall system availability to exceed 99.95%.
This implies redundancy, disaster recovery systems, and a trained workforce that will permit as
near as real time restoral as possible. New York telephone has in twenty years responded to
crises that would have sent other US corporations reeling in chaos. Their response allowed
restoral of services in extreme conditions of distress (See the recent paper by Bell discussing the
fragmentation of the telephone network).
o The focus is on operations, namely keeping the network service up to the performance
standards set, and this implies that the work force must have the capability to deal with complex
operations requirements in multiple positions. All people should be cross-trained to meet the
level of service expected by the consumer of the service.
These assumptions make the local telephone company an infrastructure entity. The telephone
company as the local operating company has a highly redundant, high quality of service, with a
highly integrated work force.
If we look at this network in terms of our architectural elements we see the following:
225 | P a g e
Control: The control is highly centralized. The control emanates from a set of methods and
procedures, flows through the overall control mechanism for the network and is integrated to the
maintenance and restoral efforts.
Interconnect: This is a truly hierarchical interconnect. It is based upon the Central Office Switch,
which is designed to conserve bandwidth at the trunk side. It provides a level of common access
based on a single voice channel.
Transport: The basic transport is the twisted pair to the end user. There is fiber in the loop and
some fiber to the user, specifically those users in the higher data rate category.
Interface: Generally the interface if the telephone handset. More recently, the interface has been
expanded to include data sets.
In contrast to the ongoing performance levels of the Regional Bell Company, the interexchange
carriers (IECs) do not have to meet the same levels of service. The recent AT&T service outage
in New York with their Class 4 and Class 3 switch outages showed how the level of service has
dropped for the IEC level (See Bell p. 36). Specifically, the IECs have recognized that a level of
service may be priced on a differential basis. This is in contrast to the pricing structures for the
local operating companies whose service levels are more closely controlled by the state Public
Service commission. It is of question if there is to be a divergence in service levels over time,
between these two types of carriers.
There has been great discussion on the threat to the local carriers from the bypass carriers. The
bypass carriers have had some impact but has not been as serious as expected. The true
competition has typically been from the changes in technology. Specifically the user has in the
past used a set of single voice circuits not having the capability to multiplex them over several
higher data rate lines. With the introduction of T1 or DS1 (1.544 Mbps) circuit, now it is possible
to get 24 voice circuits for the cost of 8.
In addition it is also possible for customer to get DS3 or 45 Mbps circuits for 8 times a DS1.
Thus the technological innovations in multiplexed circuits are the internal competitors to the
existing single voice circuits.
The economic performance of a public switched network has evolved over the past ten years in a
new direction. The rational behavior of any firm is to maximize their profits (Pindyck and
Rubenfield). This rational economic behavior has been the incentive for any company to invest.
226 | P a g e
However the PSN is inherently a monopoly in many market segments and thus operated as a
monopolist. In 1980, the strategy of the PSN provider was to maximize the rate base subject to
the constraints applied by the Public Service Commission (Kahn, Spulber).
We introduce performance elements, P i, which defines the measure that is maximized or

minimized in the rational business decision process. We can relate it to the standard utility
function of microeconomics, but will not do so in this paper. In the past, the optimization criteria
for a telephone operating as a public utility was;
P PSN = max ( Rate Base Universal Service, Level of Service)
In contrast as the regulation has been changed some new forms of performance measures have
evolved.
Specifically, we have;
P PSN = min( Cost| PSC Price Caps)
From a rational investment point of view, the corporation should be acting in such a fashion so as
to maximize the net present value of the company as reflected in its discounted cash flow. This is
the strategy that leads to maximum return for the shareholder (See Fruhan for value creation,
transfer and destruction, also Lawrence and Dyer for the analysis of the transition of AT&T from
strategy 1 to strategy 2). This optimal strategy can be given as:
P PSN = max( NPV(Cash Flow)| Level of Service)
The current evolutionary thinking in the public switched network takes it from ISDN to
Broadband ISDN with ATM switching. It is an evolutionary path that is built on the world view
of the nineteenth century system of hierarchical switches and the need for universal service. It is
driven by a need to provide the same data transport, at almost the same time, to business
customers as to the residential customer. It does not readily admit the flexibility of multiple
overlay networks, allowing for segmentation of service. It is predicated on the assumptions of
high per user capital investment and allocation of costs in a rate based world. Much of this
thought process is, however, driven by responding to a Public Service Commission demand on a
state by state basis for equality of service.
The question may be asked as to why the local operating companies have not taken a more active
role in the development of the regional networks. There are several obvious answers and those
not so obvious.
227 | P a g e
Clearly, in the case of most regional networks, NYSERNet being one, inter LATA transport is
necessary.
New York Telephone took a key role in the early stages, however it had to deal through
Rochester Telephone to legally work with the NYSERNet efforts. In fact, it is even considered in
violation of the MFJ to manage an inter LATA network. Thus, for MFJ and other regulatory
reasons, the local operating companies are wary of involvement. A second reason is the general
voice focus of the Regional Phone Companies. The emphasis on data in the networks of the type
such as NREN go beyond the general network infrastructure supported by the local operating
companies. A third reason is the questionable economic viability and extensibility of such
academic networks.
6.7 CATV NETWORKS
CATV networks have evolved over the past thirty years from a simple local distribution service
for off the air television transmissions to highly complex communications networks with a
centralized architecture.
CATV Network Structure

Strand 1:1 Strand 1:1:1
Strand 1 Town
Local Sub
Hub Hub
#1
Sub
Hub
Head End Strand 1:24:24
Switch, and Data Strand 1:24
Broadband PON
Facility Passive Or
Optical GE
Splitters
Or
Gigabit Strand 24:1
Ethernet
Sub
Hub
Town
Local Sub
Hub Hub
#24
Strand 24
Strand 24:24:24
Strand 24:24
In the early 1980s the CATV carriers were required to make all of the cable systems have the
capacity to carry two way transmissions and it was implemented in 1980 by the Warner Cable
company in the now classic QUBE system (See McGarty 1982, 1983 for a technical description
of the system and Couch, McGarty and Kahan for the market structures). Two way cable was
further expanded to provide for both data and voice transmission. The first data circuits were
228 | P a g e
provided in 1982 by Warner in Pittsburgh, providing 1.5 Mbps data communications to
Westinghouse in a metropolitan area network. In 1983, Cox Cable (see Tjaden) implemented,
with the assistance of MCI, a fully switched voice communications system over the cable
network. In 1983, Warner Cable reached 100% penetration of optical fiber in its backbone
trunks, from hub to head ends, in the large Metro systems.
In late 1983, Warner Cable developed, and field tested a full motion video, on-demand, videotex
system in a joint venture with Bell Atlantic, Bank of America, Digital Equipment Corporation
and GTE. After determining that the time was not appropriate this trial was canceled (See
McGarty and McGarty, 1983). However, GTE took the Warner developed technology and
further developed for its cable trials in California. Unfortunately the GTE trials lacked a market
driven partner and they too failed. Despite the failures, these trials showed that the cable
industry, almost a decade ago, had developed, tested and marketed systems that are still to be
implemented in the telephone network. This limit is not a technical or market limitation, it is
clearly a limit of judicial mandates on market expansion.
Thus in many ways the cable industry was far ahead in meeting the needs of all classes of users
as compared to the telephone companies. Much of this was a direct result of the lack of
regulation and willingness on the part of the customer to infer the existence of and agree to an
underlying price performance curve.
6.7.1 Structure
The CATV networks are structured to provide for the distribution of entertainment to the home.
These networks are centralized in form; they have a head end that operates the local area, a set of
hubs that are distribution points and possibly sub hubs for local regeneration. The systems are
generally one way broadcast but have a two way capability. The one way may be 50 MHz or
more and the return channel is 50 MHz or more. The systems are primarily coaxial cable,
although they are being expanded to include fiber. Some of the existing systems have a fiber
backbone The architectural elements for the CATV networks are as follows:
Control: The control is generally from a central facility in the network. The systems are
broadcast only and thus distributed control of the network is limited.
Transport: The transport is very passive. The coaxial cable used in these systems is a limiting
factor. It has repeaters along the route and these are limiters to the cable design. With the
introduction of fiber in the CATV local loop, however, this will change and the cable transport
will have the capacity for any bandwidth. Even in the current systems, there is not the
segmentation as in copper. Cable could provide any user in the reverse path with any part or all
of the reverse 50 MHz of bandwidth today. In fact, in several of the Warner QUBE systems, this
was commonly used for commercial data circuits. Thus the CATV systems are dramatically
229 | P a g e
different than Telco system in that ability to assign and allocate full capacity on transport to any
user.
Interconnect: The interconnect capability of a cable system is fully distributed. In the QUBE
system, a polling scheme was used which was centralized. In the Cox INDAX system, the
scheme was a CDMA system that was fully distributed. Thus in CATV both extremes have been
used. In fact, CATV used the first version of the IEEE 802.6 protocol in the CDMA designs, this
being the basis for SMDS systems.
Interface: Cable allowed a wide variety of interfaces, ranging from TV converters, to PCs and
video games. This is dramatically different from the limitations on Telco networks.
The cable systems have a natural monopoly in their underlying franchise licenses. These licenses
are awarded on a municipality basis and generally assure the local governments a certain percent
of the gross revenues from the system. Thus it is of immediate gain to the local governing bodies
to maximize the revenue from the cable systems to achieve the greatest revenue source from the
cable operator. The current law limits this implicit tax to 5 %. We consider the competitive
environment of the cable system under three differing scenarios and these are presented in
Appendix A. The conclusion of that analysis is that cable is a stable monopoly under current
conditions.
The original criteria for economic performance was one that was driven by returns to investors.
The CATV business was a means for many investors, using both tax shelters and capital gains
rates to obtain significant overall returns from their investment. Thus, negative cash flow in the
early years was very important to shelter income and positive cash flow deferred to later years
was essential for the payout.
This led to the following criteria for cable systems.
P CATV = max( Cash Flow Service Areas, IRR)
This has been changed to:
P CATV= min( Cost| IRR, Level of Service)
due mainly to the change in the tax laws. In addition, there has been a maturing of the cable
systems and they are now run on the basis of commonly accepted operations and business
230 | P a g e
factors. That is, the driving need for capitalization is no longer as critical. It could be questioned,
that given the great capital needs for the cable systems, if the tax law had been changed ten years
earlier, would there be the cable infrastructure that we see today. Clearly in the current finance
markets, the answer would be no. CATV systems were means to ends for many investors in the
early 1980's. The success in building a CATV infrastructure was clearly a success of a carefully
executed tax policy that allowed for the investment in potentially risky investment. The change
in that policy as clearly changed the CATV investment market.
The evolution of the CATV networks has been based on the assumption that the CATV operators
will assume the massive capital investment risks with the opportunity to attain an adequate return
on their investments. That return has for the most part, been focused on the use of the
transmission medium for the almost exclusive purpose of entertainment distribution. The fact
that the current cable business passes almost 80% of the US homes and has about 50%
penetration, thus providing service to 40% of the US households. In contrast however, even at
the current rates of service, the revenue for the CATV industry as a whole is only $18 billion,
comparable to one RBOC, there being seven, plus GTE. The total RBOCs plus GTE equal the
interexchange carriers in revenue, including AT&T, MCI, Sprint and the other IECs.
Thus the total CATV industry is about one sixteenth or 6% of the telecommunications transport
business for voice and data. Such a position is not one of any market dominance in the network
area. However, as has been noted, this industry has a network in place with both existing
capabilities and new transport capacity that greatly expands what the RBOCs have in all their
networks. The difference is driven by both the unregulated nature of cable plus the fact that cable
is not forced to be the transmission provider of last resort. That is cable is not required to provide
"Life Line" services below cost.
231 | P a g e
Network Evolution Map: Combined
Complex
Customer Customer
1995 Now
Telco CATV
1995 1995
Interface
Complexity
Telco CATV
Now Now
Private Private
Now 1995
Simple
Homogeneous Heterogeneous
CATV companies now find themselves in an environment where the RBOCs are potential
alternative carriers competing with CATV (See Carnavale). Specifically, the RBOCs have
indicated their intent in providing local loop fiber transport. That transport would have the
capability to provide transport for CATV systems. The large capital intensiveness of these
systems, however, will slow their installation. The inertia of local governments, the true
regulators and customers in this market, will probably, for the short term, maintain the status
quo. It does however allow for the CATV companies, as they progress to fiber based transports
to themselves become the alternative carrier. If the CATV companies, in the current rebuild
cycle that will occur in the 1990s, position their rebuilds in fiber, then they are in the position to
have the transport base for transmission to the home. It is then possible, in extremis, for the
RBOCs to become leasers of transport service from the CATV companies who will have fiber
passing the majority of homes, and not vice versa. It will be an issue of who will be at the home
first with the transport facility and then it will become an economic decision.
6.8 PRIVATE NETWORKS
Private networks are those offered to select portions of the general market and are owned and
operated by network operator other than the common carriers. Typical of such network operators
are Teleport in New York and Boston. These private network operators build their own bypass
capability, including cable or fiber, along rights away that they may lease from local agencies.
They then also provide some limited switching but for the most part connect to the local point of
presence of the Inter Exchange Carrier. Their selling proposition is generally lower cost for the
232 | P a g e
same service. In fact, their actual benefit may lie in the shortened time to installation and the
support on moves, adds, and changes.
6.8.1 Structure
Private networks are structured on a local basis, allowing direct access by the end user or
customer. They focus on the segment of the market that desires an additional carrier and who are
concerned about cost.
The networks generally are built to maximize the local market penetration advantage and are not
structured to provide universal service. For example, in Chicago, the Chicago Fiber Network,
was built in the coal tunnels as rights of way, directing traffic off the Michigan Avenue business
district. Teleport in New York is directed at both the financial district and the mid-town business
district.
The networks are, for the most part, fiber point-to-point and have limited any switching.
Specifically, they are built around the smart mux technology and are positioned to attract
customers on the basis of competing with bulk transport buys. For example, they sell DS0, DS1,
and DS3 circuits at significant discounts.
We can now compare the four elements of these networks to those of the others discussed.
Specifically;
Control: The control of these networks is centralized and is generally under the control of the
organization responsible for their operation. This is an organizational control philosophy and
there is generally no reason for not having a distributed control.
Transport: The transport is usually over fiber networks. In contrast to CATV, the transport is
controlled, generally by voice mux equipment and is thus no different than the Telco networks.
However, some private networks are using special data transport facilities and nonstandard data
rates. Thus it is possible to use any amount of the available bandwidth.
Interconnect: The interconnect is highly flexible and allows for all types of device interface. The
switching is done by multiplexers that are intelligent and generally are controlled by users
segments. Thus distributed control is possible.
Interface: Most interfaces are possible.
233 | P a g e
There are few examples where Private Networks compete with each other, Boston have three
smaller versions. Generally their self-competition is along common rights of way and it is yet to
be seen that there is sufficient traffic for them to survive. In Boston, as in many other cities, the
key to a Private network provider is the Right of Way. Usually the local electric, transportation,
sewer, water or gas company has existing rights of way that are usable. The problem however, is
the pricing of that right of way as well as the ease of expanding it. For example, in 1984, the
New York Metropolitan Transit Authority, MTA, was approached by several private network
providers with an offer to lease rights of way. Six years latter there is still no effective policy on
how to do it. In Boston the process took only three years. In Washington, D.C. it took eighteen
months. In Chicago, only nine months. Typical rights of way rates are $0.50 per foot per month.
That is about $2,500 per mile per month. In New York, if rights of way were available, the entire
network would need no more than 50 miles of fiber, thus costing $125,000 per month or $1.5
million per year. Such a system could supply service to several hundred customers at the level of
a gross revenue in excess of $50 million per year.
Thus the right of way is a fixed fee and if efficiently run, the network may have that fee represent
a small percentage of the total revenue base.
The Private carriers receive little if any competition from the common carriers since the common
carriers generally have higher rates commensurate with the switching, management, and
distribution services that they provide.
The approach of the private network companies is to provides for the least cost design in the
early phases subject to minimal market coverage. This financial strategy assures them the ability
to compete on a price basis during the startup phase where they capture market share based
solely on price competition.
P PVT = min (Cost Service Area) As they have evolved the criteria has changed to;
P PVT = max (Cash Flow| Service Area, Level of Service) which demonstrates the rational
approach of focusing on cash flow out of the business. As we indicated in the studies by Fruhan,
value creation is based on cash flow subject to market retention.
Private Networks have evolved on both the local and the long distance basis. On the local side,
there will be a continuing fragmentation of some of these network until the consolidation phase
occurs. The consolidation will not occur unless and until it becomes clear what the market is. As
we have indicated, the provision of service to the residential user still requires capital per user
234 | P a g e
that exceeds that for the short term payback. In contrast, the commercial user still has the short
term payback and thus will be a buyer for these services. The main evolutionary factors will be:
Market consolidation: Will there be enough clustered revenue potential for the independent
private network providers, especially if they are competing on price and in a pure transport
commodity market.
Scale consolidation: If there are no other elements other than transport, will there be scale
economies in transport on the local level. On the long distance level, such companies as
Williams have found scale economies via leveraging their gas transport facilities. Are such
consolidations available in local transport? Is there a stable environment in the franchise fee for
rights of way. As has been seen in the cable industry this has been the case on the local basis as
long as predatory pricing is not perceived? This however will not be the case in the private
network area.
Technology consolidation: As we shall note in a latter section, there are certain technologies that
may make the capital intensiveness of fiber a non-issue. Specifically radio technologies may be
capable of providing some modest bandwidth requirements.
Usage expansion: As the bandwidth is liberated in private networks, as it will be before the
public switched network, will such bandwidth drive up the revenue per fiber while at the same
time driving down the price per bit per second? It is anticipated that if the private carriers are
able to provide dark fiber and that if vendors such a startup as Ultranet and others provide high
Mbps customer based switches, the applications will increase.
Clearly, the possibility exists for the Private Network providers to cream skim the market. It has
yet to be proven that such has occurred. At present, therefore, there is limited regulation on such
networks and it is appropriate to let that continue until they have stabilized with a recognizable
character and understood economic effect in the telecommunications community.
The directions of Private networks are in the balance between the Public Switched networks and
the Customer Networks. Clearly, the former provide the network of both universality and last
resort. The latter provide the network of optimized user customizability. Noam has recognized
this dynamic and had indicated a possible direction:
" The breakdown of monopoly is due to the very success of the traditional system in advancing
telephone service and making it universal and essential (eg a need). As the system expands,
political group dynamics take place, which lead to redistribution and overexpansion. This
provides incentives...to exit sharing.. to a system of separate sub-coalitions.”
235 | P a g e
This then begs the question, if this approach is the true market dynamic, what is the stable point
of the evolutionary telecommunications architectures. Undoubtedly there is both an economic
need and social need for a Public Switched network. User sub-coalitions, namely user
empowerment, will lead to effective Customer Networks. What then is the future of the Private
Networks? Specifically, if Public networks support the wide base of users and the need for
universal service, and if customer networks provide service optimized for the larger user, what
does this leave for the private networks. It is argued here that the private networks will become
niche market players, providing a boutique service to mid-tier market users of communications
service and a broad and common functionality basis. Thus it is concluded that the private
networks will have limited broadband functionality, providing at best raw fiber transport,
marketable on a limited basis.
6.9 CUSTOMER NETWORKS
Customer networks are the newest of the five networks to have evolved. They are currently the
creature of large corporations that have the need to meet specific high density data and voice
traffic and academic consortia that are meeting the needs of their research communities. In
addition these companies have a wealth of experience in managing data and voice networks and
may even be computer companies themselves. Thus IBM and Digital are examples of companies
that have their own networks. These companies have had third parties build fiber networks that
the individual company then takes over and operates as a part of the corporate infrastructure.
These networks cover areas of heavy traffic and are priced in as cost competitive on the basis of
being less costly than leased lines. However, in most of the situations discussed the networks
also are part of the company's strategy to expand its product offering to include networking
products and expertise. Thus IBM and DEC clearly are interested in expanding their product
offerings in the network area and the most effective way to do that is through a network of their
own.
In contrast, the academic networks, such as NYSERNet and NEARNet, have focused on the
needs of their constituent members. Some of the networks use a leased circuit transport but there
are many user owned and operated fiber networks extensions. MIT, for example, has an
extensive campus wide fiber network that is used to support its local interconnect and transport
requirements. A detailed discussion of campus networks is presented in Ames.
On the commercial side, there is also the creation of shared networks for the use of market
leverage. The recent study by Kosynski and McFarlan provide several example of corporations
and their networks applications and infrastructures. In all cases they lead to a sustainable
competitive advantage.
6.9.1 Structure
236 | P a g e
A customer based network is generally a closed environment that is created solely on the basis
on providing better service at lower costs. The Customer Network may be for a single customer
or for a coalition of customers. It should be recalled that the network consists of the control,
interconnect, interface, and transport functions. The nature of a Customer network is that each of
these may be separately or jointly under the control of the customer or the customer coalition.
For example, the transport may be leased from a Public Switched Network provider or from a
Private Network provider.
The control may be directly under the customer or it may be outsourced. However, the ultimate
control is that of the customer.
The network architecture is based on purely economic factors and they are generally established
between clusters and high usage and traffic premises of large companies. The networks are built
on the following basis:
Transport: The transport is a fiber network that is built between locations. The transport is
provided on the basis of both dark fiber and a switched transmission facility common to the
switched network fabric.
Interconnect: In the corporate network, the interconnect function is provided by the inherent
point to point links and the addressing generally done by the TCP/IP transport layer protocols
that are used on the data transport. Voice interconnect is accomplished via the PBX facilities that
are on the customer premises.
Control: The control of these customer networks is based on separate network management
facilities.
Data networks are controlled by systems that are now a natural part of the computer systems.
The control of the Customer networks has taken several operational directions. Some companies
have decided that outsourcing of the control directly under their management is the way to
proceed. Other companies have maintained control under their direct corporate control. Others
such as NYSERNet have found third party companies which that had prior relationships with to
take a more active role.
Interfaces: These are provided as a common part of the end user devices.
The corporate networks allow for expansion and customization.
237 | P a g e
The competitive environment is limited in this type of network. Competition is generally from an
internal comparison with other carriers. The network is viewed as a strategic corporate asset that
is part of the overall company value chain. Thus for example, the network for Citicorp or ADP
may be part of the overall financial services system that is provided to their customers. The
choice of build versus buy is both a strategic and tactical decision. On a strategic level, the issue
is that of establishing a barrier to entry to the competition, allowing the provision of new and
innovative services. The tactical issue is that of lowest cost for a required level of performance.
The advantage of a Customer Network is the ability to optimize it for the benefits of the user.
Thus NREN, if viewed in the paradigm of the Customer Network solution, will allow for
maximization customizability in all four network elements. This is in contrast to the Private
Network which will provide a common denominator capability focused on its market niche.
The initial optimization criteria is based on the need to prove the system in on the basis on return
on investment. Specifically;
P CUST = max (Reperformance, Coverage)
As the networks have evolved the measure has changed to;
P CUST = min(Op Costs Level of Service, Functionality)
Specifically, the tactical issues of cost containment are key. Levels of service and extent of
operational functionality also play roles as constraints.
This type of network has the greatest potential for evolving. The key to this expansion is the use
of the dark fiber option allowing for the use of data rates that exceed the 45 Mbps maximum
provided by the common carriers. Companies such as IBM will undoubtedly use these networks
to expand their own products, stretching the abilities of computers to communicate at higher data
rates. The challenge is to allow the network to act as a computer backplane and reducing the
overall latency of the transport. Such low latency transport will evolve with new applications and
new computer architectures. In addition there will be a more fully distributed computer
architecture for the use of these transport facilities.
238 | P a g e
Network Evolution Map: Internet
Complex 1993
•NREN
•Multimedia
•Gbps
•LAN/MAN/WAN
1990
•2D Visual
1985 •OS, TCP/IP
•LAN/MAN
Interface
Complexity
•E Mail
•File Transfer
•LANs
1980
•Email
•TCP/IP
•DARPA Net
Simple
Homogeneous Heterogeneous
Network Infrastructure
In addition to the corporate players in the Customer Networks, the coalition players, be they
companies or academic institutions, have a significant opportunity to customize and experiment
with these networks. It is possible that it will be with the Customer Network, such as an NREN,
that the maximum amount of innovation will occur. This is clearly the case from a commercial
perspective with Customer Networks in other areas. The Securities Industry Association
Network in New York presents an opportunity for this market segment, in their coalition
structure, to establish a highly cost effective transaction processing network. This is leading to
changes in database, transaction processing computers and interface workstations.
239 | P a g e
Network Characteristics
Factor Public Switched CATV Private Customer
Ownership Regulated Telcos Unregulated Unregulated Customer
Control Telco CATV Operator Third Party Self Control
Management Telco CATV Third Party Customer or Third

Party
Interface CCITT, ISO Some standards Limited Standards Customer Specific
Interconnection Hierarchical None Limited Limited
Extent Nationwide Local Local As needed
Costs Tariffs Negotiated Negotiated Cost Based
Evolution Complex More capacity More Capacity More complex

Standard Interfaces Complex Interface Limited Service Increased Interfaces
Examples RBOCs Time Warner Teleport IBM

TCI CFO DEC
6.10 OBSERVATIONS
Networks represent the asset allocation of capital resources to meet market needs and provide the
end users with a means to maximize their utility function. In the area of commercial networks,
the network functionality goes to the heart of the value chain of the user and in turn to all those
affected by the user.
Networks play an important role in value creation not just value transfer. Networks are no longer
just intermediaries that facilitate the work of other information users, they become an integral
part of that use.
In this paper we have substantiated several key conclusions.
First: Networks will evolve in a dynamic fashion, not being driven by a single vision but being
responsive to a collection of constituencies. This will result in a collection of multiple overlay
networks.
The challenge will not be in controlling a single standard, but in developing a flexible interface
to these multiple overlay networks. Infrastructures are important, but the infrastructure is not a
physical infrastructure, it is rather a logical or relational infrastructure.
240 | P a g e
Second: The future of networks is driven by uses and users. We are entering a network
generation dominated and driven by the end user and their uses. Admittedly, the use of
regulation has been argued as necessary for the attainment of limited social goals but for the
commercial or government user, networks should be designed and evolved to meet the needs of
such business entities. Market makers are the users.
They will define the boundaries of their networks in terms of the uses to which it is applied.
Thus, the users will seek networks that are enabling and not delimiting. User communities are
not homogeneous as in the classic paradigm for hierarchical networks. It is because of that
paradigm shift and the needs to meet niche interests that the change is occurring.
The end user community is not going to wait for an ultimate infrastructure. That is an economic
reality.
The end user will be empowered by a new architecture and this architecture will result in sets of
quasi-independent multiple overlay networks.
Third: Current regulation is based upon past paradigms. Networks are evolving and creating new
paradigms. In this changing world view, regulation serves to buffer obsolescence but should be
carefully controlled not to endanger the facilitation of value creation and transfer. Multiple
overlay networking is the essence of true competition. If the goal is to maximize the benefit for
the consumers and to avoid the monopolistic bottleneck, this is best accomplished by fostering
the ability of users to have the networks of their choice.
Fourth: Technology is a fluid change agent in the stability conditions of networking. The free
market drivers of minimum cost and maximum utility impact will assist to integrate the new
technologies into the alternative network solutions. The technological lifetime of new network
architectures may be significantly less that in the past. The capital intensiveness of networks are
changing as a result of both existing infrastructure and technological changes. The mindset of the
rate based regulation and barriers to entry due to capital investment may no longer be valid. This
will force a change in the view of networks as distinct logical or relational infrastructures as
distinct from a national highway physical infrastructure.
A Research Network and in some sense LANs are distributed suboptimal consequences of the
continuing asymmetrical regulation rather than independently successful technological changes
for end user application. If regulatory policy would respond to the changing world view, if it
would allow for the implementation of less controlled broadband systems, unfettered all players,
optimal solutions are more likely. These more efficient solution would further be in the best
economic interests of the user.
241 | P a g e
This paper has viewed networks not as a national physical infrastructure, as some have argued, a
parallel to the national highway system. Rather, it is argued that networks are flexible and less
capital intensive logical or relational infrastructure. NREN is a concept that fits that view and
could potentially provide the basis for future users and use development. However, NREN as
conceived and to be implemented should be nothing more than an academic test bed which may
have the potential to evolve into a limited operational network for use limited to researchers.
NREN is not and cannot be of any value to a commercial set of entities because of its
fundamental lack of value creation end user support.
In addition, technology changes in networks at rates that dwarf the highway model. Roads are
built today in a fashion similar to the Romans of two thousand years ago. Networks built today
are dramatically different than those of ten years ago. Thus the paradigm that states that networks
are the next highway infrastructure may lead to disastrous results if implemented. It will result in
the strategy of least common denominator, the most intensive capital costs, and the greatest
regulation. In contrast, if we view networks as fluid creativity and value creation for the end
user, then it is essential that successful economic development be left in the most open market
possible.
6.11 CATV MARKET DYNAMICS
In this appendix we develop the details of the argument that leads to the conclusion that the
current CATV market is an established monopoly. We proceed by considering the set of cases
that may be considered in the evolution of competition in this market. The demonstration relies
on an understanding of Game Theoretic analyses in econometric systems. The presentation is
non-mathematical. Consider, now the following cases;
Case 1: Franchise Licenses (Exclusive):
The exclusive franchise license is inherently a stable monopoly (See Shubik, 1987 for a
description of the applications towards game theoretic approaches). It allows for a maximization
of the cable franchise fee by providing a maximum value to the underlying asset. The only
competition that the CATV provider would then have is the alternative displacement competitors
such as home video, off the air television, and the movie theaters. Generally, the pay as you buy
alternatives such as rentals and movie theaters are incremental and more discretionary, whereas
CATV is considered a fixed monthly expense. Thus there is an inherent bias on the part of the
consumer, based on the standard consumer inertia factor, to rank order the cable purchase as the
first choice. There is also, as has been shown in the past few years, been an inelastic demand for
cable services, almost independent of price. The response of the local governments has been de
minimus as compared to Federal Congressional bodies. The local governments have seen their
local revenue increase almost 350% in the past five years.
242 | P a g e
Case 2: Non Exclusive Franchise, No Alternative Incremental Transport:
In this environment we assume that each CATV provider must build their own transport network
and that the franchises are nonexclusive. We will argue that the result of this is again a stable
solution which is again monopolistic, assuming that there is an existing cable provider,
established as the dominant. If there are no dominant providers, we will argue that the stable
solution is that no CATV provider will either bid or survive.
It should be first noted that the value of the franchise is less in this scenario to the franchise
holder.
Non exclusivity has reduced the potential revenue to the local government, so that this and the
third scenario have lower marginal returns to the local government.
Let us now consider the two such classes of this case:
Prior Dominant Carrier:
Assume that there is a prior dominant carrier, one who has the existing franchise and that the
build of the system has been completed and penetration in excess of a minimum level has been
achieved. In this case, the customers have been captured, they are accustomed to the service and
have adapted to the monthly fee.
Further assume that the service provided, namely the basic service channels and the pay
channels are generally available to any CATV provider on a comparable basis. This is a key
assumption, it assumes that there is no differentiator of the service based on the product. Namely,
that all CATV products, that is the video content, is comparably the same.
Let us now proceed with the market dynamics under this scenario. It will be noted that this
scenarios has the structure of a two person zero sum game. Thus, from an analytical point of
view, the dynamics are those that satisfy the Nash criteria (see Shubik, 1984, p. 194 and Luce
and Raiffa p. 140; also see the original work by von Neumann and Morgenstern).
We assume that the underlying competitive factor is price, since all providers have equal access
to all products.
We assume that the dominant carrier has already established the system at an average cost less
than any average cost of a subsequent carrier, and that the marginal cost per new subscriber for
the established dominant carrier is smaller than for any new provider.
243 | P a g e
We assume that the established carrier has raised per monthly rates for several cost increase
cycles so that the marginal revenue significantly exceeds the marginal costs and capital
requirements. Thus the existing carrier has at least a positive marginal cash flow.
A new competitor enters the market. The new carrier must build a new system and must "buy"
the new subscribers from the existing system or from the pool of non-subscribers. The only
economic way to do this is with lower prices.
The established carrier then responds with lower prices, still sustaining at least positive marginal
cash flow. The new carrier, due to the entry cost structure, and the fact that it has considerable
higher marginal cost structures, is at a negative marginal cash flow. This cyclic price war
escalates until the newcomer is forced from the market.
The final stable point is again with the single player, who can again raise the rates and wait for
another competitor.
The net result for the local municipality is lower franchise fees.
No Prior Dominant Carrier:
In this case, we can readily reach a conclusion based upon the results of the prior case. Since
competition is based upon price, there is a stable solution if and only if there is a clear
segmentation of the market by geography that allows for the reduction of prices to a level that
meets positive marginal cash flow. Again, this solution assures only a minimal franchise fee for
the local municipality.
Case 3: Non Exclusive Franchise, Alternative Incremental Transport:
Let us assume that no CATV provider must build their underlying network because it is provided
by a third party. Thus there is limited capital investment provided and that transport is provided
by a third party, such as the local telephone exchange company. Let us further assume that the
transport can be provided at a cost that depends upon only each new subscriber an is available on
the market on an equal price basis. Thus there is no competitive advantage on any transport
provisioning. Further there is no competitive advantage on the provision of the video product.
Thus the only controllable costs of each CATV provider is that of its local sales and operations
organizations, all else being fixed. There are then limited scale economies available and thus the
marginal costs generally equal the average costs.
Competition in this case is based upon price and the perceived quality of service. It can be
shown that this case generates the most cost competitive market and that the franchise fees are
244 | P a g e
minimal to the municipalities. However, it may be argued that the consumer may get the lowest
priced service, the switching costs from one provider to another being minimal.
switching costs from one provider to another being minimal.
245 | P a g e
7 COMMUNICATIONS ENVIRONMENT
The ability to interconnect many users with multimedia applications has developed over the
years into a structured and layered methodology. This chapter presents to the reader an overview
of the communications environment, focusing on broadband systems, and expands to coverage of
the overall communications lawyering protocols.
The key factor that drives the communications environment for the multimedia user is the need
for a connection based service. Connection based services assure the user that there is a
synchronous service provided that allows for the end to end integrity of the image.
In this chapter, we discuss the communications element from the aspect of what is needed to
effect the overall sessions service and to implement the complete multimedia communications
environment. We take an approach that is the inverse from that of the standard communications
study, starting with the higher layers of functionality first and moving downward towards the
transport function as the latter element.
We view communications as a means to an end, but a means that has a cost in terms of delays,
capacity limitations and errors. As indicated, we begin by assuming that the communications
medium has all of the elements that are necessary for the errorless interconnectivity of all the
users in the session. It is the
assumption of all of the players being in the same room at the same time. As we relax these
assumptions, we get farther and farther into the limitations of a real communications
environment.
7.1 REQUIREMENTS AND ARCHITECTURE
The issues of the communications environment revolve around the questions of what
communications do and how much is needed to effect the necessary task. The multimedia
environment places many more requirements on a communications system than is placed by any
other usage. In the multimedia environment we are dealing with very high data rate requirements
that demand high throughput and this moves quickly into the 100 Mbps to Gbps range of data
speeds. In addition we are dealing with an environment that demands that requires that there be
minimal delays and that the messages are synchronized. Also we demand that the users be
allowed to develop the session structure that we have been developing and that the sessions be
transparent to the communications network.
In the multimedia environment, we are dealing with a world of communications that is quite
unlike that of the computer to computer communications world. In the computer only world, the
end user, namely the computer, can be designed through its software to adjust for the vagaries of
246 | P a g e
the communications channel. In fact, the development of the layered architecture that we
discuss in the next section was a direct result that the communications channel was unreliable
from the point of view of
throughput, errors and delay. This then required layer upon layer of protocols to adjust for the
channel. As we evolve new channels, we can envision correcting many of the errors of the past
and providing a much more enabling environment.
7.2 MM COMMUNICATIONS ENVIRONMENT
The multimedia communications environment is complex in that it demands that any set of users
of the system be allowed to be interconnected in any fashion and further that they share the
resources of any of the elements of the overall system, either separately or in concert. We have
seen that in a multimedia environment that we combine together the forms of communications
such as voice, video, image, text, and standard record files, and that we can create many virtual
users, such users being real people, applications programs or database and other devices. The
ultimate embodiment of a multimedia communications environment is the session which reflects
the ability of all of these users to share complex multimedia data objects in a seamless fashion.
When we look at the communications environment of the real world we see that it is often
lacking in its ability to provide the resources necessary to effect this capability. The standard
communications system is a voiced based network that provides data rates typically of 56 Kbps
or less to the end user. Even with the introduction of the proposed ISDN network, this moves up
to 64 Kbps, which is still too little too late for the multimedia environment.
There is, however, an available set of higher data rates in the larger network, rates up to 45
Mbps. However, these are not provided on a shared switched basis, only on a dedicated basis.
The question may be, why do we need shared and switched. The answer is that we really do not
want to pay for dedicated fiber from one location to all other locations on any session connection
that we may desire to connect. The shared switched network works well for voice
communications, since that is the essence of the current telecommunications network.
The network depicted is that of the shared switched public network which is generally a
hierarchical network, consisting of local switches and a higher switch hierarchy.
The environment to support multimedia communications needs is generally quite limited at the
present time. For example, at the time of the writing of this book (1989), there are only 45
45Mbps circuits that have been sold under tariff in New York City. In contrast, the city is the
media capital of the world, with the highest concentration of media intensives users anywhere.
Prime amongst these users is the printing, publishing and advertising industries. Millions of
images and hundreds of thousands of hours of video are created and moved from one location to
247 | P a g e
another around the city on a daily basis. In fact, the most common form of communications is the
use of bicycle messenger, each moving at significant speed across the city streets. In fact, these
bicycles are moving in the Gbps range!
Thus despite the needs for this multimedia communications market, there is yet an established
infrastructure. This is generally because of the lack of the needs of the multimedia
communications environment. These requirements are as follows;
1. Effective and Available I/O Devices

2. Shared Switched Networks
3. Layered Architectures
4. Protocols and Standards
5. Connection Based Services
6. Video, Voice, Data Integration
7. Local and Long
8. Distance Interconnect
Based upon the discussion of the multimedia environment it is possible to develop a detailed set
of requirements for a general multimedia environment. In this section, we focus on some general
requirements issues, but note that special network requirements may be necessary for very
specific applications.
7.3 PERFORMANCE ISSUES
As we design a multimedia environment, we are required to develop a set of measures that detail
the performance of the network. The performance measures take the loading on the network,
based upon the number and types of users and their frequency of use, and determine how well
the network can provide the services required. The issue of Grade of Service is a set of
benchmark levels that tell the user what to expect from the network. Performance tells the users
how the network can vary from the specified grade of service.
The main performance issues to be developed in this chapter are those that have been accepted as
the most effective measures of network performance. These general performance measures are;
1. Effective Data Rate

2. Throughput
3. Delay
4. Errors
5. Delay
6. Capacity
7. Variability
248 | P a g e
7.4 SIZING ISSUES
Sizing is the complement of performance. In the performance model, we assumed that the load
on the network was determined and that we then proceeded to develop measures of performance.
In the sizing analysis, we assume that the performance level have been given and we then
determine how large a user base is supportable. In performing a sizing analysis, we determine the
limits of usage of the system.
The major sizing elements for a multimedia system are as follows:
1. Number of Users
2. Number of messages per source per unit time o Message Mix
3. Maximum Data Rate per Channel
4. Maximum Number of Instantaneous Users
7.5 ARCHITECTURES
We have continuously, in this book, addressed the issues of architecture. The architecture issues
are the first in any system design and provide the blueprint for the development of any detailed
design. The architecture elements are those that must be configured to meet the need that are
articulated by the users of the system. In this section, we first detail the architecture elements and
present some of the architectural alternatives that we shall be developing in this chapter for the
communications element.
There are five architectural element that we consider important for the communications
environment. They are;
Network Topology or Physical Layout
Addressing Formalism
Switching Capabilities and Alternatives o Functionality Interfacing o Interconnection
(i) Network Topologies
We depict several network topologies that are in common use. The tree topology is a common
topology in such areas as cable television networks. This topology has advantages for networks
where the major function is the distribution or broadcasting of high bandwidth data. Thus they
find common use in cable networks (see McGarty). The tree network is a hierarchical design that
assumes that the system flows both to the headend location of the communications network. The
249 | P a g e
existing telephone network is of this type in certain locations and is clearly the case in the long
distance communications configuration.
The star topologies has been accepted as a viable local topology for fiber networks where the
fiber is generally a single mode type and cannot be adequately tapped and bridged at multiple
locations. Single mode fiber is not amenable to these types of taps and thus it usually finds itself
in point to point interconnects. To expand a point to point connection to a network topology, we
find the star architecture as the appropriate one.
The ring topology provides a totally connected structure that allows each users to be
interconnected onto the network at their location and further allows for the condition that if the
network is severed at any location, there may still be a path to other locations, albeit not
functioning under a ring structure. Thus, rings are used for applications where there is a need for
some form or redundancy. We can further extend the ring structure to include a dual ring
structure, one ring carrying traffic in one direction and the other ring carrying traffic in the
opposite direction. We shall see that this becomes a common architecture in the metropolitan
area network applications.
The bus network topology is a ring that has been broken. The bus may be a dual bus architecture
that is similar to the ring in that one bus can handle traffic in one direction and the second bus
handles traffic in the opposite direction.
(ii) Addressing Formalism
Addressing is a fundamental issue in communications networks. We frequently forget how easy

it is to communicate in a telephone network because we know how to reach someone by means
of their name. However if we look at the process in further detail, we see that the human naming
convention is a complex data structure that we have had years of experience with and have
accepted almost as part of our cultural history. We know that to call someone, we must know
first where they live, we must then find the area code from an independent source, and then and
only then can we dial xxx-555-1212. At that point, assuming that there is no telephone strike, we
ask information for the number of John Doe. We may further have to delimit this if there are
sever entries of that name. This delimitation is the location of Mr. Doe's house. Consider now
that we must do this for all of the elements in the multimedia communications network.
Addressing thus applies to users, terminals, layer control elements, data bases and anything else,
logical or physical that we wish to address. We shall see in this chapter that the essence of
communications is not only getting the bits there but being able to say where there is. That is the
addressing issue. We can look at addressing in two dimensions, shared and private networks. In a
shared network, we must all agree as to how we are to be addressed. There are limited Vanity
■ ■J
250 | P a g e
phone numbers, mostly limited to 800-xxx-xxxx. The classic number is the trouble reporting
number for N.Y. Telephone for data circuits, 800-1 AM-DEAD!
In private networks, the addressing methodologies are somewhat freer and depend basically upon
the limitations of the supplier of the hardware and software. However, as anyone who has run a
private network knows, the data and network administration task is significant.
In any shared network, the sharing is done by some form of switching. The function of switching
is to allow any user to get to any other user and in addition to allow the user to gain access, on an
demand basis, to any set of the network resources. The switching may be done in three different
ways.
Centralized: Each user accesses a centralized and hierarchical controlling element that
determines the resources necessary and determines the path that is to be taken in providing the
connection. The voice telephone network is an example of this type of switching architecture.
The signaling to the central switch may be in-band or out of band. An inband signal is one that
uses the same channel that the message transport occurs in as is done in the telephone network as
today. Out of band signaling is increasing and is typical of the type of signaling in the new
Signaling system 7 that will be a part of the ISDN network.
Localized: In this form of switching, the end user equipment takes full responsibility for the
selection, management, setting up, and control of the communications channel. Generally this is
a complex function and there are very few system that implement this strategy.
Distributed: This is a common switching strategy is its most ^common implementation is in the
area of local area networks. It provides for any users to interface with any other user by all users
watching message that flow across the network and determining which of the messages are for
them and then selecting the message. The advantages of this type of system is that it allows for
many users to activity participate in the network but it has the disadvantage of having a higher
data overhead.
Interfacing the many elements in a communications network is generally done through a layered
architecture. We shall see in this chapter, that the layered approach has been developed in such a
fashion that it allows for effective and universal interconnection of all of the elements of a data
communications network. The need for multimedia communications has not yet allowed for the
development of an equivalent set of layers for multimedia environments.
The interfacing using layered protocols makes several assumptions about the communications
environment. First it assumes that there are well defined and structured level that can be
segmented into separate control and implementation levels. Thus if the is a level that performs
251 | P a g e
just electronic functions we can segment all of these function into that level. This is what is done
in the OSI physical layer. If there is an agglomeration of functions that handle the switching
functions or the graphics interface functions, we could just as easily load all of these elements
into that layer.
Thus layering requires a commonality of functions, a collection of those functions into separate
layers, and the interfacing of those layers into the overall network. The advantage of the layered
architectures is that that they permit the development of software and systems that can be
segmented to improvements on a layered by layer basis. The major disadvantage is that with
more layers, and more segmentation, we give up speed and performance for the flexibility to
develop the individual layers. We shall see in this chapter some of the effects of layering.
(v) Interconnection
Interconnecting networks is a critical design and architectural function that must be considered in
any communications network design. The interconnection is driven by the fact that no single
network is self-sufficient and that the world is becoming an amalgam of many different types of
networks. Overlays of broadband multimedia networks are typical and will become a more
common feature.
There are conceptually four generic types of interconnections. They are:
(i)Repeater: This provides for a physical to physical interface connection between to different
networks. Thus a repeater may be nothing more than a modulation and demodulation pair of
devices in a network.
(ii)Bridge: This type of interconnection takes the first step in interconnection networks where the
data format of the bits may be different. Thus if we want to connect one type of LAN to another,
or even a LAN to a MAN, we can do so through a bridge, The bridge matches the physical
difference in the signal structure as well as the data format.
(iii)Router: This element of the interconnection hierarchies allows for not only the physical and
bit format matching but also handles the changes in the addressing between the two networks.
Thus we can go from one network where the addresses may be in one format to another network
with a different addressing scheme.
(iv) Gateway: When everything is different from one network to another, we need a gateway to
match all of the changes from one to the other
7.6 LAYERS AND STANDARDS
252 | P a g e
There has evolved a standard way to view the communications environment over the past ten
years that involves the use of layered protocols. As the reader may recall, we introduced the
layered approach first in Chapter one when we introduced the overall multimedia environment.
We extended this to software layering in the when we developed the source and in when we
integrated the GUI concepts with such elements as X windows. We further extended this in
with the data storage and file elements of the design.
In this chapter, the layered constructs have taken even a firmer form with the introduction of
international standards. This section reviews the seven layer international standards for
communications systems. Layered approaches are useful for both implementation and interfacing
purposes. In the implementations context, the layered approach allows for the clear delimitation
of functions and shows how they can be partitioned to allow for simpler system design and
coding.
Communications networking, like any form of human communications, needs to have a set of
protocols or agreed to sets of standards of handling many talkers at the same time. The protocols
are ways in which the speakers or participants in any conversation can know how to interrupt and
who has precedence in the conversation. A typical example is a meeting in a highly structured
company. If the CEO is at the table, all conversation may generally defer to that persons
comments and the level of interrupts may depend on
the title and status of the people at the conference table. Inhuman conversations, we frequently
have learned these protocols in a cultural environment and they are generally not written down.
The need for addressing written protocols occurs when we are dealing in cross cultural
environments such as a U.S. business man doing business in Japan, in that case, the difference in
and acceptance of protocols is critical.
The same situation is prevalent in communications, and more importantly in multimedia

communications. We approach this problem with the development of protocols and these
protocols are developed in a highly structured manner. Historically, these protocols were
developed in an environment of data communications, where the need was to allow for the
interconnection of computers with other computers. The protocols were not envisioned as a way
for human interaction with the communications environment. We shall see how this has evolved
as we describe the protocols in some detail.
The protocols that we shall describe are in seven layers, each of the layers having the
responsibility for a specific set of functions that are necessary for the communications network.
Our approach to these protocols is different than most in that we shall begin at the highest layer
and work downward. The typical pattern of development and the actual path of development in
the standards bodies was from the bottom up since they were interested in protocols that satisfied
253 | P a g e
the needs of the data communications community. In our analysis, we are interested, first , in
the needs of the multimedia communications user.
Thus, we develop the seven layer protocol set from the top layers down. This is a drastically
different course in the development and exposition of the OSI protocol set but it reflects the need
in the multimedia environment to have the higher layer functions come to the fore.
The seven protocol layers are as follows:
7.6.1 Layer 7: Applications:
The applications layer provides for a wide variety of services that support the overall set of
applications that may be running on the end users CPU. These applications can generally run on
a standalone basis with no need for the applications layer to even exist. It is necessary, though,
when we desire to network the CPUs together in some communicating fashion. A typical
example of an applications layer function is the file server function that is an integral part of a
local area network environment. In this applications service, a particular applications program,
such as a word processor, any desire to obtain a file of letters that have been stored on the file
server. The applications program then evokes the file server service and this places a request
from the program, through the local area network to the target file server.
Other applications services are such functions as a mail service, a directory service for all the
users , a virtual terminal service that allows users to integrate a variety of terminals on the
network, and the graphics interface services that we had developed in . Thus the applications
layer, is the highest layer that functions to allow for applications to take full advantage of the fact
that they can communicate with a variety of other users or applications on the network.
7.6.2 Layer 6 Presentation:
This layer is probably the most improperly named for the purpose of this book. It does not deal
with presentation in any way other than possibly how it "presents" data to the applications layer.
It has been suggested by Tannenbaum that it be called the representation layer, and by others that
it be called the data packing layer. It actually provides a useful function in how the data is to be
packaged and repackaged as it comes and goes on the communications network. One service of
the presentation layer is that of end to end encryption. This service takes the information that is
generated by the end user in an application, and secures it through some encryption scheme that
reduces the probability that some nefarious interceptor may capture the data for illegal purposes.
7.6.3 Layer 5 Session:
254 | P a g e
This layer is the one that the standards bodies have spent the least amount of time on but as we
have seen in this book, represents the heart of a multimedia communications environment.
The session is the construct that we have developed that allows many users, in a multimedia
environment, to be assured of a seamless, error proof path that ensures the synchronicity of the
communications across the path and the graceful interconnection and elimination of any set of
users as required. In the multimedia environment, the session is an enduring dialogue between
one or many human users and their interactions with applications programs and multimedia data
files. Voice and video do not suffer the vagaries of a connectionless network well and require
that there be a connection based entity to support their communications requirements, (see
Kishino et al, Nomura et al, Verbiest and Pinnoo, and Kishimoto et al) .
In these first three layers we have dealt with concepts that ensure that the higher level
communications processes are properly handled. We have implicitly assumed that all other
things have been handled in an appropriate fashion. The lower layers will worry about such
things as the end to end integrity of the bits, the routing of the bits to the correct places and the
physical interfaces being correct.
7.6.4 Layer 4 Transport:
The transport layer is the first "real" communications layer in that it relates to bits and bytes
going across the network. It is concerned about the end to end communications of those packets
of information, assuring that they get from point A to point B. It assumes that if it send a packet
to the lower layers that they, on a one by one basis get where they are supposed to in an errorless
fashion, but that this layer must be concerned with the whole set of packets generated by a
source.
7.6.5 Layer 3 Network:
The network layer is an integral part of a network that has many parts going in many directions.
This may be called a set of subnets. These subnets have to be controlled, and packets that rattle
around in this environment must be managed carefully and assured that they get, on a packet to
packet basis, from one network node to another.
7.6.6 Layer 2 Data Link:
The data link layer worries about bits getting from one local location to another in an errorless
fashion. This layer provides for error detection and management and may also include the
capability to correct for errors that occur on the channel. It assumes that the lower layer provides
only a transmission path for bits but that the transmission path may have errors.
255 | P a g e
A significant part of the Data Link layer is the Media Access Control (MAC) sublayer. The
MAC layer provides for the local control and access to a distributed and shared communications
facility. Typical amongst the MAC capabilities are local and metropolitan area networks.
7.6.7 Layer 1 Physical:
The physical layer includes the physical connection, the modulation scheme and the multiplexing
or switching schemes that can be used. Many of the elements of the proposed ISDN (Integrated
Services Digital Network) can be found at this layer. The common RS-232 interface is also an
element of this layer.
We can explain them in some further detail by considering a specific example and how each of
the layers interacts and requires the services of the layers below. The basic transfer element is a
data packet and that each of the layers adds a header onto the data packet to effect the services
that are provided by that layer. This goes all the way down to the physical layer which is the
layer that deals with bits.
Consider the example of a set of multimedia users that desire to communicate in a multimedia
mail environment. Let us first assume that thy are using a specific application program to analyze
a particular image on the screen. This may be the case of a single user at a single location. That
user now desires to share that image with several other users and comment on the results in some
multimedia fashion.
We generate a data block that is the multimedia image element. This is the first step in the
ultimate process. The data block is in actuality a complex multimedia data object, encompassing
video, voice and still image. At the higher layers, the notation of a data block is merely symbolic
of a large data element that is to be transmitted across the complex network.
We then invoke the mail service at the applications layer and append to the data block that we
wish sent a mail header. This header will contain all the elements that are necessary to evoke the
mail function and to ensure that the system sends this total image from one location to all other
users in the system. I will further assume that the act of mailing is such that the image being sent
is a compound multimedia object, consisting of video, voice and still image, and thus we must
have a fully synchronized session based service in place. We evoke the mail service in an
applications layer program that provides for electronic mail. It may have also been possible that
this mail program may have been resident at another location on another processor. We shall deal
with these options at another time.
Since this is a highly proprietary message, we will need to encrypt this message and this service
of the presentation layer is invoked and it is represented by the header that is appended on the
applications header packet at the presentation layer. This header is a symbolic method of
256 | P a g e
demonstrating the process of encryption. We evoke the encryption at a point in the system called
the data encrypting algorithm, which may be a standalone piece of secure hardware in the
communications system.
At the session layer we append a header to the encrypted message to send to a session manager
processor. This processor may be a network based element that handles all of the control for the
network. This element sends out a set of commands to ensure that a temporary session is
established and that all of the session participants are brought onto the session connection. The
session manager then continues to clear the channel and to assure that the ultimate block of data
associated with this multimedia message is properly handled as a complex multimedia object,
and not just a standalone computer record which can be packetized in any random fashion.
The transport layer will now take this packaged and session secure pact and it is required to get it
from one point to another, and that is all. It will assume that the data channels are errorless and
that they are being routed to the proper locations.
The transport layer takes the session packet and may break it up into smaller packets for
transmission and further takes instructions from the session layer so as to ensure the
synchronicity of the total packet as it moves across the network. The transport layer control is
located in a transport protocol control device that may be part of the computer front end or as a
standalone device.
The network layer is needed because we are transmitting this broken up, packets into a network
composed of many smaller switching elements. The network layer functions at each of these
nodes in such a way to route the packets in the most efficient form from one point to another.
This layer focuses on getting the smaller packets across the disparate network.
The data link layer deals with the error control protocol that ensures that the data is kept error
free. In this case we assume that the DLC protocol is an HDLC (High Level Data Link Control)
protocol that allows for acknowledgments and error detection. Error correction is not provided
on transmitted packet and is achieved only by packet retransmission.
Finally, the physical layer performs this task with an RS-232 interface to my computer using a
simple on-off 9.6 Kbps modem. This is the easiest layer to understands.
We call view the message at the top most layer as M and then view the actions of each layer as
Lk where k represents the layer. At the top most layer we have:
R7 = L7 M
257 | P a g e
where R7 is the layer 7 representation of the message M that is a result of layer seven operator
L7. Thus the mail function may operate by L7 on the message M to generate R7.
In general we can say that;
Rk+1 = Lk+1 Rk
Rk+1 = Lk+l Lk ... L7 M
For example, the operator LI is a modulation operator on R6. Thus R7 is a modulated waveform
with specific characteristics.
There are many issues that arise in the discussion of the OSI protocol stacks. Some of them
are as follows;
Addressing: How do we know how to get from one point to another in a network. Addresses are
necessary at all layers of the protocol.
Each layer has elements, we shall call them entities, that are agents for performing tasks at those
layers. The RS-232 interface or the modem are physical entities, and the mail process is an
applications entity. We must be able to identify entities at all of the layers.
Active agents are necessary at the layers to effect the operations of the services provided at those
layers. These are called service access point (SAP) and these are the points that effect the service
operation. We can usually envision these being at the interface between the two layers.
There must be an agreed to format and set of rules for the layers to exchange the data downward.
This format is called the interface data unit (IDU).
The IDU consists of a set of control information called the interface control information (Id)
and the actual information to be passed along the network, called the service data unit (SDU).
There must be a set of calls or instructions that allow for communications at the same levels in
different units in the network. We shall call these primitives and they have their own
syntax. The syntax of the communication language primitives fall into a form :
PRIMITIVE = LAYER.Service_Type.action(parameters)
258 | P a g e
For example, at the session layer we may be asking the session SAP to create a session and then
to add several people to the session. The primitive may take the form:
Session.Create_Session.request(pi,p2pn)
This sends a request to the session processor that a session creation is desired. We can further
extend the syntax by including parameter within the primitive to delineate the details of the
primitive request. For example:
SESSION.Session_Create.request(source_address,destination_address)
Thus the general form for a primitive will be;
LAYER.Service_Type.action(ul,u2,....,un)
We show the elements of the layers and show how this primitive form of interaction within
layers functions
There are four types of actions that can be implemented in the different protocol layers. These
actions are;
1. request: asks an entity to do something
2. indication: entity is informed of a request
3. response: entity desires to respond
4. confirm: entity is made aware of request response
The type of Service_Types may vary by layer.
We shall be spending significant effort on the Service Types development of the primitives,
especially the Service_Types. These are extremely important at the higher layers for the
implementation of the services that can be implemented in these types of networks. We have
already discussed several primitive concepts in Chapters 2 and 3 in both the graphics and
windowing areas already. The development of these primitives in the communications area will
merely be an extension of that effort.
We can envision the two entities in a communications mode, one being the Initiator and the other
the Responder. There are two layers of protocol involved. At the Initiator, the user at the higher
259 | P a g e
layer send a request action to their lower layer. This goes across the network and becomes an
indication passed up from the lower layer (Provider) to the comparable layer (User) who then
responds. This response is called the response and it goes down and up and it results in a confirm
from the Initiators Provider layer to the User layer (see Schwartz p. 88).
7.6.8 Applications Layer
The applications layer is the highest layer and it is structured to support the end user applications
as well as provide an environment for the applications to reside in. The applications layer's
functions, as we shall describe shortly, focus on direct support of the applications programs. The
applications programs may have access to the applications layer services through the primitive
functions that we discussed above. It is through these primitives that we obtain access form one
layer service to another.
We may view the applications layer services as if they were adjuncts to the operating system, in
that we frequently call the operating system services from pure applications program.
7.7 FUNCTIONS
7.7.1 Applications
The applications layer functions both as a service provider as well as a port to the lower layers of
the OSI stack. In its function as a service entity, it directly provides services that utilize the
network resources that are available to it. Thus, the applications layer can provide such network
based services as;
Mail
File Server o Directory
The applications layer may also provide for locally resident services that may be of general use
to the end user. These types of services are those of the graphics user interface, terminal
emulation services and other specific translations of generic to specific nature. We shall focus on
the functionality of the applications layer the relates to its use of the network resources.
7.7.2 Presentation Layer
The presentation layer is in actuality a data re-presentation layer. It takes the data that is provided
by the session layer or provided to the session layer and performs certain tasks to represent the
data. For example, the presentation layer takes a letter that may be generated in the mail service
at the applications layer and encrypts the letter using an encryption service. The service layer
260 | P a g e
may also take an image that is to be sent at the applications layer and may compress the image
using a data compression algorithm of the types that we have discussed in . This layer may also
provide a means for changing from
one data format to another, especially in terms of how the data is to be presented to the
applications layer.
The presentation layer does not deal with presentation of information to the end user. That,
ironically, is dealt with in the applications layer as an applications service. The presentation layer
deals with the presentation of data to the applications layer and to the session layer. It focuses on
how to change the data from and into forms that meet the overall communications requirements.
Also recall that the data or information at the higher layers is not just a block of data it is the
entire message.
5.2.2.1 Functions
The presentation layer has several major functions as we have been discussing them.
Tannenbaum has presented them as follows;
Provides an access mechanism to the session level services.
Provides a common mechanism for the definition and manipulation of complex data structures.
Provides a management facility for the management of the data structures that are currently in
use through the applications or session layer services.
Providing a means to convert a data element from one form to another, on either transmit or
receive, for the purposes of more effective or secure communications.
We can see that these types of functions become more complex when we deal with the
multimedia environment. Recall, that in the multimedia environment that the session layer tasks
on a more significant role and that this role requires that the images be handled in a more robust
fashion. The present conceptual structure of the presentation layer deals with data structures
that have little time variation. That is they are considered give as a block of data in their entirety
and do not have the synchronization problems that we face in video or voice communications.
We can envision, possible, the use of presentation layer services that can be used to create the
multimedia data elements, and to provide these to the session layer.
Consider, for example, that there is an applications layer program that generates a complex
multimedia object. Let us call that object DOk. We let DOk be given by;
261 | P a g e
DOk = {DOVk,DOIk,DODk,DOTk}
where these sub elements are those for voice, image, video and text respectively. We may
envision that the presentation layer has the responsibility to handle the manipulation and
integration of this multimedia data object. We have seen in how we can generate this object in a
single data file and we shall further see in how we can handle these elements in distributed
multimedia database. We see in the presentation layer a functionality that allows us to
manipulate these data elements across the network so that we have a mechanism to maintain
their total integrity.
7.7.3 Session Layer
The session is the heart of the multimedia communication environment. All that the layers below
the session layer do is guarantee that the data packets have gotten from one point to all of the
other points that are in the session. They do not guarantee that they got there on time or in most
cases even in order. In addition, these lower layers of the protocol sets only view the data
transported as a single block, even if it is a created complex multimedia data object. However,
the data object may be a sequenced element of an extended session between many users and the
context that that block of data finds itself in is as important as the block in and of itself. The
session layer is structured to deal with communication "in context". For the most part, it is the
only layer that deals with the interactive dialogue nature of conversation environments as a
natural part of it functional and service elements.
If we have an error free fiber connection between each and every user, it is conceivable that
there is no need for the lower four layers and the only critical layer is that which maintains the
session. Ironically, the session layer was not considered as an important layer in the early days of
the ISO structures. Tannenbaum discusses the fact that the British had originally proposed that
the session layer was not at all needed. As we shall see in this section, the session layer provides
some of the most valuable functions from the perspective of a multi user multimedia
environment. All the session layer assumes from the services below it is an error free path for the
messages that is sends to those layers. All it expects from the layers above is a continual flow of
messages that are to be forwarded to those users. The session layer takes the responsibility of
managing the interuser conversation, not just the communications.
5.2.3.1 Functions
The functions of the session layer relate to the maintenance and support of the session services
that we have just discussed. Some of these functions are as follows:
262 | P a g e
Dialog Management: This provides for the control over who is intruder control of the session at
any one time. We can view the management of a dialogue handled in a variety of forms. The
baton approach is based upon a mutual agreement between the session participants that the
holder of the baton is in charge and may exercise direct control. The baton may be passed back
and forth amongst the session members. Another approach is a priority based session dialogue
manager which assigns a priority level to each of the session participants. We shall discuss these
and other latter in this section. There are performance issue that relate to deadlock of dialogue
management and the delay in session control.
Session Establishment: Session establishment allows for one user to initiate a session and to add
other users onto the session in an operational context. The session establishment function allows
for the identification and addressing of other virtual users and ensures that they are bound to
these session. It also allows for the disestablishment of sessions. We frequently are concerned
about the session establishment time as we are concerned about call set up time for a telephone
call. Thus in terms of the overall performance of the session establishment function, we look at
determining the overall session establishment set up time budget and allocate it to the various
system resources used.
Connection Based Service: A session is synonymous with a session. A connection based service
is one that establishes a permanent virtual circuit that the session is built upon. The
establishment, maintenance and disestablishment of a session is one of the functions of the
session layer. The session is built upon the session layer created connection based circuit, that is
the permanent virtual circuit that the session layer creates with the transport layer.
Synchronization: When we are dealing in multimedia communications, we are dealing with

information that is not only contiguous in terms of spatial relationships but also has a temporal
relationship that is critical to its presentation, display and interaction. We may think of what may
happen to a video display using an NTSC scan line system if instead of 525 lines being
transmitted we get only 524. This would propagate down to the bottom of the system and result
in a totally confused interaction. Errors and transport faults do occur and the timing of messages
across the network may face significant delays and temporal variances. It is the function of the
synchronization service of the session layer to keep all of these complex messages in a truly
synchronous mode. The suynchronization service performs this task by inserting various
synchronization ticks in the data stream and recovering them at the other locations in the session.
Activity Management: A session is an enduring communications between two or more entities in

the communications network. There are moments when the actual communications in the session
may be reduced to a low to zero level. We can envision breaking up a session into elements that
represent the burst of active interaction between the users and during those times allocate
services of the lower layers. We will always want to have the session endure, since the endurance
263 | P a g e
is a quality of the session. However, it is a function of the session layer to recognize the dormant
phases that may occur in a session and to allocate resources accordingly. We do this by creating
or defining elements of the total session called activities that are the periods of the active
participation of the users of the session and during those periods allocates the resources of the
system. This session service allows for the optimum use of the communications resources at the
lower layers and assists in the optimization of the shared switched services for transport.
5.2.3.2 Protocols and Architectures
The session layer protocols are similar to those at the layers above but focus on the
implementation of the services described in the functions that we presented in the preceding
subsection. Recall that the syntax for is as follows:
LAYER.Service_Type.activity
Here we have SESSION for the LAYERS and we still have the four possible activities as:
request indication, response, confirm.
These actions relate to the development and specification of the protocol to effect the service.
The Service is the indication of which of the services that is being invoked. These Services are:
Session, Connection, Dialog, Activity, Synchronization, Report
The Type sub-indicator may tell that it is the beginning, end, discard, resume or interrupt of an
activity service. The session Type delimiter provides additional specificity to the command.
Recall also that part of the syntax are specifiers to the command syntax the indicate the specifics
of the session elements. We have detailed several detailed primitives that are used in an actual
multimedia system.
5.2.3.4 Alternatives
Let us consider several implementations of the session layer functionality. The first deal with
those of dialogue management and then we consider synchronization and activity management as
well as exception reporting.
In dialogue management, the major function is to develop algorithms that allow for the
management and control of sessions. This management and control can be handled from the
point of view of a token or baton or from a wide variety of other techniques. Let us consider four
that are common.
264 | P a g e
Hierarchical: In this case a session manager in control at all times and all requests for the baton
are passed up to the session manager. The requests are handled in some priority fashion based on
equal user priority and first come first served to some other set of priorities. It is a non-
distributed control algorithm and requires that each session select and adhere to the single point
of control.
Baton Passing: This is a round robin approach where the baton is held by a user for a fixed
period of time if they need it otherwise it is passed to another users in a round robin fashion, each
users is allowed a maximum period to maintain session control and at the end of the period must
relinquish it. It may be relinquished earlier.
Priority: In this case, all of the users have a priority, and no two users have the same priority. The
priority may be a combination of several factors. Let P(k) be the priority of user k, and we define
this as:
P(k) = R(k) + T(k) + D(k) + I(k)
where ;
R is the hierarchical level of rank of k
T is the length of time since k last controlled the session
D is the data waiting in k's buffer
I is the image intensity of k's session
This is a totally arbitrary priority scheme but any other priority function may be generated. Th
scheme works by all users sending out their priority to the network and the users comparing
theirs to others and responding accordingly. The only users to respond is the one with the
greatest priority.
Interruption Control: This scheme is a first come first served, shouts the loudest protocol. It is
based upon a user's sending a broadcast message out to all users demand access and only when
this users messages are received first by all others is this session accepted.
7.7.4 Transport Layer
The transport layer provides for end to end integrity of the communications path. It assumes that
there is an errorless and routed infrastructure in the lower layers and ensures that the message
that is to be transmitted is done so errorless in toto. For example, the transport layer will receive
265 | P a g e
a mail message from the higher layers and its takes the entire mail message, partitions it in a
fashion to meet the limitations of the transmission paths, and sequences smaller packets across
the network to ensure that the total mail piece is sent in an errorless fashion. It is important to
note, however, that though this performance may suffice for a computer oriented system, it is not
sufficient for the multimedia environment developed in this book. The transport layer allows for
the transport on an end to end basis of a message. The session layer ensures the endurance of the
conversational mode that is the essence of the multimedia environment. The session layers
assumes that the transport layer exists. The transport layer makes similar assumptions of the
layers below it.
5.2.4.1 Functions
Transport layer services can be of the connection based type of the connectionless type. The
connection based type transport services follow the requirements that we placed on them in the
session layer. In this case the transport layer may either support the session layer or may even
supplant it in the session
connection service. The connectionless type transport service is merely the ability to provide for
a datagram service, ensuring that a single package is transmitted in and error free fashion from
one point to another.
Recall that at the transport layer the services are provided to it from the Network SAP and are
provided to the session layer from the T SAP. The actual hardware and software that performs
the tasks in the transport layer are at the Transport Entity.
The typical functions of the transport layer are:
Connection Establishment: This function provides for the calls down to the lower levels for the
establishment of a connection between another user on the network. It is a connection which may
also be a permanent virtual circuit and in some ways may emulate the session level connection
service. If the transport layer can support a full connection service, then the session service is
simply provided by a call to the transport service.
Addressing: Addressing at the transport layer provides for the naming and identification of
specific point to be communicated with in the total network. At the transport layer, the
addressing function allows for specific identity of the physical and logical location of the virtual
user.
Connection Disestablishment: This service is the termination of the connection established in the
connection establishment.
266 | P a g e
Connection Management: The management of a connection is necessary both from the
perspective of maintaining the connection as well as providing reports on the status of the
connection as it progresses in time. This service allows for interface for the overall network
management of the services.
Flow Control and Buffering: The information handed to the transport layer from the higher
layers is always the total message and not a fragmented form that is suitable for communications.
The transmission paths generally do not support the transport of large blocks of data because the
errors inherent in raw transmission would never permit an effective communications system.
Thus the communications path uses some form of smaller packets, usually a small percentage of
the total information package. Thus the transport layer must accommodate the buffering and
segmentation of the total information element and control the flow of the packetized versions
across the network.
Recovery: Recovery is a key service at the transport level and it is used as an element of the
overall session service. Recovery provides for the restoration and reconstitution of the
communications path is there is a fatal failure that causes the collapse of the path. We generally
assume that this function is at the transport layer since it has full control over the total network
elements.
Level of Service (Quality of Service): The transport level allows for the attachment of level,
grades or qualities of
service to the established communications path. These grades of service are related to the
performance factors that we shall be discussing in the latter parts of this chapter.
Multiplexing: The transport layer can provide a service that allows for multiple circuits on the
same channel by the proper identification of sources and sinks in the network. We shall expand
upon this latter.
These services can be created in a fashion similar to those of the higher layers. In the next
subsection, we shall focus on some of the specific primitives and how they are architects to
provide the transport layer functions or services.
7.7.5 Network Layer
This layer provides for the point to point switching and routing of packets through the network.
The services that the data link layer provides is an errorless point to point path. The network
layer has to provide the transport layer with a clear path from the beginning to the end of the
transmission medium.
267 | P a g e
5.2.5.1 Functions
The functions of the network layer are as follows: Routing, Congestion and Flow Control,
Internetworking, Protocols and Architectures, Performance Issues.
7.7.6 Data Link Layer
The data link layer has traditionally had the responsibility of providing an error free transport on
a point to point basis. It assumes that there is an underlying physical connection, but that the
connection may be error prone. It further assumes that the connection has been made correctly in
some gross sense by commands that have been sent down from possible higher layers in the
protocol stack.
There is also a second subfunction that is now incorporated in the DLC layer and it is in the
Media Access Control (MAC) sublayer. This layer provides for interconnection of physical
devices in an errorless fashion on a distributed network such as a local area network or a
metropolitan area network.
7.7.7 Physical Layer
The physical layer describes the interfaces that the computer communications system has with
the real world of the communications network. It discusses the physical interfaces, the
modulation schemes and signal levels, the data rates and coding schemes, and even the switching
and multiplexing schemes. We shall not focus a great deal on this layer in this book since there is
a wealth of literature on this layer elsewhere. The reader is referred to such books as
Tannenbaum, Schwartz, and others.
There are several functions of the physical layer. They are; Physical Interface, Mechanical to
Electrical Conversion, Switching, Transmission, Modulation, Multiplexing, Protocols and
Architectures, Performance Issues.
7.8 BROADBAND ALTERNATIVES
The most pressing requirement for the communications system in order to ensure the ubiquity of
the service that we are developing in this book are the elements that relate to the transport of data
in very high data rates, these rates are in the regions of tens of Mbps to the higher Gbps range.
These rates are now achievable with the introduction of fiber technology, fibers that are made of
glass filaments. These fibers have been developed and perfected in the past ten years and now
are at the stage that they are both ubiquitously deployed and in addition various
implementations are available for implementation.
268 | P a g e
In this section we discuss three of the most common alternatives for the deployment of high
Mbps data channels, ones that also have a potential for evolving into Gbps channels. From a
technical point of view, a single strand of fiber can theoretically support data rates in the Tbps
range, that is 10 12 bits per second. However, the technology to interface this data rate to the
network is not yet available in commercial form. We shall not discuss these factors in this
chapter but leave this to the reader. The texts by Personick, Kaiser and others provide significant
detail on the fiber technology and the status of the opto electronic interfaces.
The choice of the three techniques discussed in this section is based upon the facts that they are
either available of will be currently available in the near future. FDDI is a presently available
system that supports 100 Mbps data channels and is being deployed for intra premise network
systems. It frequently works on multimode fiber systems and can be interconnected to inter
premise systems. SMDS is an embodiment of a metropolitan area network architecture that is
being planned for shared switched network deployment. It works at the 45 Mbps rate and can go
to the 150 Mbps range. It is a dual bus architecture and can support a shared switched network
operation. Finally, ATM is the embodiment of the Broadband ISDN system and is farther off in
implementation.
7.8.1 FDDI
We start first with a broadband system that is presently in operation and provides for 100 Mbps
capability in internetting local area networks. The FDDI(Fiber Distributed Data Interfaces)
systems are typically used in intra facility networks although they are not limited to these areas.
The generally are also used in an non switched and non-shared environment, although changes
are being developed to permit them further capability.
The FDDI systems were developed to meet the need of the users of LAN systems and to provide
the capability to allow these system to be interconnected at higher speeds. Initially, the LAN to
LAN interfaces were at considerable lower speeds and thus made for slow file access and
transfer time on these types of networks, recognizing the needs for expanded interfaces speeds
the use of 100 Mbps was the desired bit rate on the interconnect network. There was, however,
no thought originally given to the implementation or operational issue of a multimedia
environment. We shall however show how this can be incorporated into this environment.
The architecture of the FDDI system is built upon a dual ring bus structure that uses a fiber optic
cable. The fiber cable is frequently a multimode fiber and the bus architecture allows for graceful
degradation of the system in the event a single node or a fiber failure occurs. The ring
architecture has been chosen in place of the other forms because of the inherent reliability of the
ring in the event of a single point failure. Each node has a repeater function as well as the
capability to detect the FDDI packets as they are transmitted around the network.
269 | P a g e
The IEEE 802.5 MAC protocol is used on the ring for the control of the access to multiple users.
This protocol is a token passing scheme that passes a token around the network that allows the
users to grab the token in the event that they have data to transmit or otherwise to pass the token
to other users. As we discussed in the previous section, on the topic of the 802 standards, this
particular IEEE standard allows for control on ring structures. The other protocols are generally
found on bus structures.
the message an indicate whether it has been received or not properly and retransmit this change
message around the circuit.
The message will continue to transmit around the ring, and when received by the original station,
it will read the message and will recognize if the message was successfully transmitted or not. If
not a retransmission may be performed if it has not been timed out.
Let us first consider the implications of the circulation of this single token. The speed of
propagation of a signal in a fiber is approximately 200m/microsec and the free space speed of
light is 300 m/microsecond. Thus the speed of propagation of about 0.66 that of free space. Now
consider a ring message packet of 100 bits and consider an FDDI ring of 1,000 m. The ring
propagation time for a packet to go around the entire ring, assuming that there are no delays
anywhere, is 5 microseconds. If the message is 100 bits, then in order for the bits not to overlap
the data rate must be greater than 100 bits/5 microseconds or 20 Mbps.
When the token bit is changed, that station now can send data for a period as long as THT (token
holding time) seconds, usually 10 msec, and then it must release the token. When a station
desires to seize the next token, it does so by looking at the data message being sent around the
ring and when it wants to seize the token, it enters a message in the AC byte to indicate its
priority. If there is a message there already, it waits. If there is a request there already but it has a
higher priority, then it may be able to bump the other request. When the station stops due either
to time out of data finished, the token is passed to the requesting station if there is one or a blank
token request starts circulating again.
The performance of the FDDI networks has been studied in some detail and we present the result
outlined in Bux. Some of the analysis of this protocol for two different channel rates and for the
presentation of the mean transfer delay time as a function of the information throughput. Several
observations are clear;
The greater the length of the cable the greater the delay.
The greater the length of the packet the greater the delay.
270 | P a g e
There is an instability that occurs in the delay as the throughput attempts to reach the data rate of
the bus. Thus the maximum throughput is less than the maximum bus rate as is expected.
FDDI applications are quite appropriate for intrafacility locations or those involving shorter
distances. As the distances increase we have seen that the average delay increases
proportionately and that this is an inherent element of the token ring structure.
7.8.2 MDS
This represents a second step in the development of a broadband capability. SMDS, switched
multimegabit data services, is a shared switched metropolitan area network (MAN) that is being
proposed for implementation in local transport areas. It is built around the 45 Mbps standard.
The architecture for the SMDS system is comprised of a dual bus structure, not a closed ring, and
the forward part of the bus transmits packets with data plus a bit that is called the busy bit. The
busy bit indicates if the packet is empty or has data. At each node or station on the SMDS
network, the packet is decoded and the busy bit is read. We shall discuss the status of the busy
bit shortly.
The SMDS system is a connectionless datagram approach to higher speed communications

networks. The network designs are developed for the use in metropolitan area networks (MAN).
There is a second or reverse bus in the dual bus scheme. The reverse bus transmits a packet
containing no data but it does contain a request bit (REQ). This bit is generated by any one of the
stations upwards of the receiving station. The bit is set to one by the station desiring to transmit.
The stations read both the forward and the backward bits.
The SMDS protocol acts in the following fashion. Recall from the previous paragraph that on the
forward channel a station that transmits data transmits not only the data but a bust bit, BUSY.
We shall call the complement of this the IDLE bit. If the IDLE bit is l then the packet is empty.
Also recall that on the reverse channel we have the request bit, REQ, that if it is 1, it means that
some station upstream desires to transmit.
We can now describe the SMDS MAC protocol for transmission purposes. This is called the
Distributed Queue Dual Bus (DQDB) algorithm and is the heart of the 802.6 MAC protocol.
Each station has a Request Counter, called RQ. The contents of the RQ at any time k is called;
RQ(k) = contents of the request counter at time k

271 | P a g e
When a data block on the reverse channel desires to send a message it generates a REQ bit that is
loaded into the RQ. Thus when the station sees a reverse packet with a REQ we have;
RQ(k+l) = RQ(k) + REQ(k)
When the forward channel sees and IDLE bit set at time k+2, it subtracts the IDLE bit from the
request counter, RQ.
RQ(k+2) = RQ(k+l) - IDLE(k+l)
Thus RQ measures the number of requests left upstream less the number of idle packets that can
satisfy those requests. Or, in
effect, RQ is a measure of the status of the upstream demand at any time.
When a node wishes to transmit it transfers the contents of the RQ to another buffer called the
countdown buffer, CD.
It then separates the RQ from the forward path and keeps it connected to the reverse. Thus RQ(k)
becomes;
RQ(k) = RQ(k-l) + REQ(k)
The countdown buffer is connected to the forward path. Then the CD buffer is as follows:
CD(k) = RQ(k); at the time of transfer
CD(k) = CD(k-l) - IDLE(k); when an empty packet passes
When CD(n) = 0, the station transmits a packet.
This protocol is dramatically different than the FDDI 802.5 protocol since it incorporate a
minimum distributed intelligence in each node of the network.
The SMDS protocol has three layers. They are;
Level 3: Provide the SMDS addressing information. Also error detection is handled at this level.
272 | P a g e
Level 2: Segmentation and reassembly is provided at this level.
Level 1: This is the physical layer functionality.
Some simple performance analysis has been done on the SMDS protocol. To data, however,
there is not as detailed a study of the DBDQ protocol as there is of the many other protocols.
However, Newman et al have presented an analysis of the DBDQ protocol in terms of the access
time delay and compared it to FDDI. As we noted in the FDI case, the delay increase as the
network increase in size and also as we increase the traffic. In fact, it is possible to make the
FDDI delay dependent upon the distance of the ring alone and have it independent of load. This
fact is not at all evident.
SMDS provides an intermediate choice for data communications over a larger area, typically
100km and greater. It allows for many users to access a network and to do so at speeds that are
consistent with both the telephone network and he local area network speeds. Thus SMDS
presents an ideal candidate for metropolitan area networks (MANs).
7.8.3 ATM/BBISDN
The previous two networks are in various stages of implementation with their standards having
been well defined and developed and the technology either available as with FDDI or clearly
under development as is the case with SMDS. Broadband ISDN, is in some sense the son or
grandson of ISDN which is slowly having limited introduction in some telephone areas. The
broadband version of ISDN has two major directions. The first is the development of a
synchronous system which has been built around the SONET standard and the second is an
asynchronous system called the asynchronous transfer mode (AT) and this provides the most
effective transport environment for the full end to end transport.
7.9 NETWORK MANAGEMENT
The issue of network management addresses the need of both the operator of the system to
control the resources available and the need of the system user to obtain maximum benefit of the
system and all of its resources. The task of network management is best stated by what is has to
do for the user of the system. A good network management system must allow the user who is
employing a wide set of system resources, and when the system fails in any way to meet the
user's needs, such as the return key not work, allowing the user to have a single source of contact
who will return all of the resource to the user in a form which allow the user to complete their
tasks. This function must be achieved in as transparent and direct fashion as possible.
In this section, we discuss network management from the perspective of the communications
system, but we must remember that all of the elements of the multimedia environment must be
273 | P a g e
controlled in as seamless a fashion. This network management in a multimedia environment
must deal with the end user from the interface, through the data and file structures, through the
communications network and including all of the network services and applications. This is a
goal of such a system and this goal may not be readily achievable in the context of many of
today's systems. However, if we begin to recognize the need of this function, we can more
readily factor these into the overall design at the earliest possible stages. This is an example of a
typical multimedia environment that needs the network management problem. Specifically, the
network includes the following elements;
The network management problem can be stated as is shown. We have shown several of the sub
network elements and have in addition shown that the elements are comprised not only of their
physical interfaces but of all seven OSI layers. The problem then is if there is a network problem
in any one layer it must be identified and corrected, and more importantly, if a problem arises as
a result of the interaction between layers, then this is the critical issue associated with network
management.
The solution of the network management problem is to develop a manager of managers that
looks at all of the seven layers of each of the sub network management elements in each. It then
can present an integrated system to identify, correct and restore the network functionality.
7.9.1 Functions
The network management system must perform the following functions;
Interfacing: This function allows for the interfacing of the management system to talk to and
control other sub network management systems. The interfacing function can be approached in
two different ways. The first fashion is the approach of defining a standard interface to the
overall management system and have all of the sub network vendors meet this interface. The
second approach is to have a flexible and referable interface that does not require vendor
modification.
Status Monitoring: This function provides for the monitoring of the status of each of the
individual sub network elements.
Performance Monitoring: With the results from the status monitoring the performance
monitoring function provides for an evaluation of how each of the sub network elements are
functioning with regards to the standard in which they are to perform.
Performance Analysis: The function of analysis determines the details of the system errors and
faults and allows for the determination of how and where this can be improved.
274 | P a g e
Inventory Management: This function is a key ingredient that provides for a ready access to all
inventory of other elements of the system as well as the ability to achieve the restoral that is
necessary.
Restoral Activation: This function provides for the reconstruction of the assets of the network to
perform its overall tasks.
Reconfiguration Management: This function provides for the management of the overall
reconfiguration task in contrast to the restoral activation function that actually implements the
restoral. The reconfiguration management, looks not only at the resole functions of getting the
system back to a prior state after a fault occurrence but also, and more importantly getting the
system into a new state, one that is planned and orchestrated.
Report Generation: A necessary function is the preparation of reports that are both event and
time driven in nature. An event driven report is one that occurs when a particular event occurs. A
time driven report is prepared at specified time intervals. The report generation function of the
network manager must be flexible in its ability to meet the changes in the environment as
well as flexible to be presented on various out display devices, print and video.
7.9.2 Architecture
There are many subnetwork elements in the system and there are multiple interfaces in the
overall architecture. The overall system architecture shows that the system operates in a multi
domain applications. The system must cover two dimensions of operations. The first dimension
is the interfaces across multiple sub network management systems, typically integrated at the
lower level of the OSI, layers. The second level is within the same network and upwards
thorough the OSI layers. A third dimension is desired but highly complex, is that integrating the
first two dimensions.
The overall architecture consists of two major elements; the software and the hardware elements.
The system architecture shows that there are common interfaces between the overall network
manager and the sub network elements, these interfaces may be of a customizable type or even
matched directly to each interface element. The software of the system must provide for the
flexible interfaces to each of the systems and then must also be able to provide for the functions
that we have defined above. The display of the results as also a key element in the design of the
network management capability. The display function may have to be flexible to meet the needs
of all the end users.
7.9.2.1 Software
275 | P a g e
The system software architecture is depicted below. It is composed of multiple layers that
internally match the interfaces to the local subnetwok managers and externally face the end user.
The inner facility of the network manager is the kernel functionality that must meet and interface
with the multitude of other subnetwok elements. The outer layer is called the shell and it is
through the functionality of this layer that the end users have access to the system.
We further take the functionality at each of the two layers and further divide it into foreground or
real-time activities and background or nonreal time activities. The need for this partitioning is
both for the implementation of the codes as well as for the implementation of the manager on a
real time platform or target machine.
The kernel functions include the following:
Communications Interface: This function provides for the subnetwok interface and handles the
protocol support to manage the vagaries of the different interfaces. This function allows for the
parsing of incoming messages and the same function for outgoing commands. The general
276 | P a g e
parsing function provides for a flexible command interface and allows for the operation of that
interface as a common element in the network manager.
Performance Monitor: This is a time and event driven real time function that monitors all of the
incoming status reports and determines if the report indicates an error status. The error status can
be divided into several categories;
Error Level Detection: This provides for the determination of the severity of the error and
categorizes it instantly for action.
Error Correlation: This is a complex task that takes less severe errors or faults and generates an
assessment of the interaction of these faults to generate more complex errors.
1. Performance Analysis
2. Status Monitoring
3. Report Generation
4. System Analysis
In a similar fashion, the shell level functions can be given as:
1. Performance Management
2. Fault management
3. Configuration Management
4. Security Management
5. Accounting Management
7.9.2.2 Hardware
The network management hardware architecture is fairly simple. The network is a complex
collection of many elements and the network managers function is to acquire from a wide set of
existing interfaces a complete collection of the network faults. These faults are locally generated
and they are collected either in band or out of band of the existing signal paths. The in band
collection uses the existing network paths during the periods in which they are properly
functioning. In the event that they cease functioning a separate set of communications paths must
be established. These are the out of band paths. Some architectures have these out of band paths
in place for all communications. The latter approach may be costly and also occur the risk that
they themselves may be error prone.
277 | P a g e
7.9.2.3 Implementations
This section describes several of the network management system implementations that are
currently available.
7.10 NETWORK PERFORMANCE AND SIZING
The issues of performance and sizing are complimentary issues. Performance addresses the
issues of; given a specific load on the system and a specific set of system capabilities, what are
the performance factors for the system in such areas as delay, throughput, response time and
others. Sizing, on the other hand, addresses the issue of; given the fixed set of performance goals
in terms of capacity, delay, throughput etc, how many users can the system support and how
large should the system resource should be. Thus it is clear that when we address on of the two
issues we address both.
In this section, we address the two issues from the viewpoint of the overall multimedia
environment and do not provide a detailed discussion of either of the two elements. We leave
these as details to be developed in the references and in the problems the end of the chapter.
The communications element of the multimedia environment is frequently thought of as the most
easily understood part of the overall system. All that is needed is faster bandwidth an greater
connectivity. It is however the mosey complex because of the time scale of communications
network evolution and system design. There are communications network that are intra premises
and those that are inter premise. In this chapter we have developed the concepts of the standard
architectures that we all assume will take care of the environments and their mixes. However, we
see that there is clearly an asset of multiple alternative for the multimedia environment even at
the lower layers of the protocol.
In this chapter we have also introduced the concept of network management. This is a critically
important element in the overall multimedia system design and must be integrated into all of the
elements that we have developed in this book. Unfortunately this is not the case and the user is
often let to fend for themselves in the event of errors or faults occurring in some part of the
network.
We have also further developed the clear distinction of sizing and performance in this
environment and we shall extend it to the overlap environment of the multimedia network as we
progress.
278 | P a g e
8 SESSION MANAGEMENT
Multimedia Communications involves itself with the communications of highly distributed

multimedia data objects that require precise timing at and between multiple locations. This paper
proposes a way to handle this level of communications through enhancements made at the Session
Layer of the OSI protocol standard. The approach taken starts with a definition of multimedia data
objects and then develops the required elements for the Session Layer. Detailed implementations
are presented and discussions on their performance comparisons are discussed. Multimedia
Communications is a discipline that combines the ideas of the human senses, disparate storage and
data structures, varying interfaces and complex communications systems. The basic concept of a
multimedia environment has evolved from that of the single media data focused world of the
computer specialist to the need to provide a fully integrated system for a human user to interact
with using information stored on many different storage media. Multimedia consists of a matching
of the three elements of the senses, the storage media and the interface devices.
It has been argued elsewhere that multimedia should not be confined to merely the storage of
information of multiple storage devices. Rather, multimedia must include the senses and the
interfaces as well. In fact, for the purpose of this paper we define multimedia as the confluence of
storage, senses and interfaces. Specifically, multimedia relates to constructs of not only
information storage but also information processing and communications. It encompasses all of
the senses, although we currently only focus on the senses of sight, sound and touch. The definition
that we take of multimedia in this paper is an expansive definition. It has been taken to provide a
basis for the next step which is multimedia communications, which takes the multimedia paradigm
and adds multiple human elements and as such transcends the prototypical computer
communications view of the world.
When we introduce the communications concepts, we do so in the context of having multiple users
share in the use of the multimedia objects. Thus multimedia communications requires that multiple
human users have sensory interfaces to multiple versions of complex objects stored on multiple
storage media. In contrast to data communications in the computer domain, where humans are a
secondary after thought, and optimization is made in accordance with the machine to machine
connection, multimedia a communications is a human to many other human communications
process that must fully integrate the end user into the environment. Multimedia communications
thus generates a sense of conversationality, it is sustainable over longer periods, and it has an
extreme fluidity of interaction.
Various authors have recently addressed the issue of multimedia communications with an
architectural approach. (See Little and Ghafoor, Nicolaou, and Steinmetz). The current approaches
279 | P a g e
focus on one of two extremes, either on broadband communications and the transport mechanism
or on the multimedia storage aspects of the system design. Little and Ghafoor have attempted to
integrate the presentation and data object side of the problem and have at a higher level, attempted
to address the communications issues. Nicolaou has developed a communications architecture that
follows the OSI standards but in attempting to introduce the multimedia issues has been forced to
introduce several new constructs. Various other researchers in this area have focused on the lower
protocol layers and have specifically been concerned with transport layer problems and below.
One of the major challenges to multimedia communications is that today there are broadband
architectures that are developed that provide higher speed communications using direct extensions
of the techniques developed in the data world of packet communications. Specifically such
techniques as ATM and SMDS, as well as FDDI are direct offshoots of local area networks and
packet technology. The fail to understand the paradigm that we are developing in this paper that
relates to the structure of the multimedia object and the conversationality of multimedia
communications.
In this paper, we concentrate on three issues in the area of multimedia communications; the data
objects, the conversationality of the interaction and the overall communications architecture. We
first note that the data structures in multimedia environments are dramatically different than those
in normal computer data communications. Specifically, Mullender has shown that typical data file
sizes that are transferred in a UNIX environment are on the order of 2K bits whereas in a
multimedia environment the file size may average 100 Mbits. Secondly, a multimedia environment
needs to handle real time data interaction such as that in real time voice and video. As is well
known, such transport protocols as TCP/IP are not adequate from a delay perspective to support
these types of data objects.
The conversationality aspect of the multimedia environment is key to effective communications.

In this paper we focus on utilizing the Session layer from the OSI format for the delivery of the
multi-user conversationality. Historically, the session layer (See Tannenbaum) has been relegated
to a secondary position in the OSI hierarchy. In a multimedia environment, we show that the
session functionality, refined and expanded, provides the essential integrating capability for
conversationality.
The remaining communications services, at OSI layer 4 and below, become, at best, delimiting
factors in the communications environment. In this paper we show that there are certain underlying
performance factors of the lower four layers, that when combined control the overall end to end
performance as viewed from the users perspective. As a major point in this paper, we argue that
the standard approach to communications system design, from the physical layer and up is the
wrong way to proceed for multimedia. Specifically, in a multimedia environment, one must,
perforce of user acceptance, design the system from the top layers and down.
280 | P a g e
8.1 MULTIMEDIA DATA OBJECTS
In a more standard computer communications environment, the data objects have significant
structure and they are frequently integrated into a system wide data base management system that
ensures the overall integrity of the data structures. In a multimedia environment, the data elements
are more complex, taking the form of video, voice, text, images and may be real time in nature or
can be gathered from a stored environment. More importantly, the separate data objects may
combined into more complex forms so that the users may want to create new objects by
concatenating several simpler objects into a complex whole. Thus we can conceive of a set of three
objects composed of an image, a voice annotation and a pointer motion annotating the voice
annotation. The combination of all three of these can also be viewed as a single identifiable
multimedia object.
Before commencing on the issues of communications, it is necessary to understand the data objects
that are to be communicated. We can consider a multimedia data object to be composed of several
related multimedia data objects which are a voice segment, an image and a pointer movement (e.g.
mouse movement). As we have just described, these can be combined into a more complex object.
We call the initial objects Simple Multimedia Objects (SMOs) and the combination of several a
Compound Multimedia Object (CMO). In general a multimedia communications process involves
one or multiple SMOs and possibly several CMOs.
The SMO contains two headers that are to be defined and a long data sting. The data string we call
a Basic Multimedia Object (BMO). There may be two types of BMOs. The first type we call a
segmented BMO or SG:BMO. It has a definite length in data bits and may result from either a
stored data record or from a generated record that has a natural data length such as a single image
screen or text record. We show the SMO.
Figure: SMO Structure
281 | P a g e
Synch Decomp BMO: Basic Media Object
The second type of BMO is a streamed BMO, ST:BMO. This BMO has an a priori undetermined
duration. Thus it may be a real time voice or video segment.
A simple multimedia object, SMO, is a BMO with two additional fields; a Synchronization field
(Synch) and a Decomposition field (Decomp). We now depict the SMO structure in detail. The
Synch field details the inherent internal timing information relative to the BMO. For example it
may contain the information on the sample rate, the sample density and the other internal temporal
structure of the object. It will be a useful field in the overall end to end timing in the network.
The second field is called the Decomp field and it is used to characterize the logical and spatial
structure of the data object. Thus it may contain the information on a text object as to where the
paragraphs, sentences, or words are, or in an image object, where the parts of the image are located
in the data field.
These fields are part of an overall architecture requirement finds it necessary to provide an "out-
of-band" signaling scheme for the identification of object structure. The object structure is
abstracted from the object itself and becomes an input element to the overall communications
environment. Other schemes use in-band signaling which imbeds the signal information with the
object in the data stream. This is generally an unacceptable approach for this type of environment.
When we combine these objects together we can create a compound multimedia object. A CMO
has two headers, the Orchestration header and the Concatenation header. The Orchestration header
describes the temporal relationship between the SMOs and ensures that they are not only
individually synchronized but also they are jointly orchestrated. The orchestration concept has also
been introduced by Nicolaou. In this paper we further extend the orchestration function beyond
282 | P a g e
that of Nicolaou. The concatenation function provides a description of the logical and spatial
relationships amongst the SMOs.
Figure: CMO Structure
O C CMO

O: Orchestration
C: Concatenation
These concepts have been further developed in McGarty[2] and there we have provided more
detailed structure to the multimedia data objects. We then also plot the times that the CMO, the
concatenation of all simultaneous objects, change in this system. Then we also show the CMO
headers that are flowing in the network at each change interval. It is this dynamic process of data
elements that must be controlled by the session layer to be discussed in the next session.
283 | P a g e
Figure: Temporal Interaction of CMOs
CMO Headers
Changes
Video
Voice
Image
Time
We can also expand the concept of a CMO as a data construct that is created and managed by
multiple users at multiple locations. In this construct we have demonstrated that N users can
create a CMO by entering multiple SMOs into the overall CMO structure.
The objectives of the communications system are thus focused on meeting the interaction between
users who are communicating with CMOs. Specifically we must be able to perform the following
tasks:
Allow any user to create an SMO and a CMO.
Allow any user or set of users to share, store, or modify a CMO.
Ensure that the user to user communications preserves the temporal, logical and spatial
relationships between all CMOs at all users at all times.
Provide an environment to define, manage and monitor the overall activity.
Provide for an environment to monitor, manage and restore all services in the event of system
failures or degradation.
We shall see in the next section that the session layer service address all of these requirements.
284 | P a g e
8.2 SESSION LAYER FUNCTIONS
The OSI layered communications architecture has evolved to manage and support the distributed
communications environment across error prone communications channels. It is presented in detail
in either Tannenbaum or Stallings. A great deal of effort has been spent on developing and
implementing protocols to support these channel requirements. Layer 7 provides for the
applications interface and generally support such applications as file, mail and directory. The
requirements of a multimedia environment are best met by focusing on layer 5, the session layer
whose overall function is to ensure the end to end integrity of the applications that are being
supported.
Some authors (See Couloris and Dollimore or Mullender) indicate that the session function is
merely to support virtual connections between pairs of processes. Mullender specifically deals
with the session function in the context of the inter-process communications (IPC). In the context
of the multimedia object requirements of the previous section, we can further extend the concept
of the session service to provide for IPC functionality at the applications layer and specifically
with regards to multimedia applications and their imbedded objects.
The services provided by the session layer fall into four categories:
Dialog Management: This function provides all of the users with the ability to control, on
a local basis as well as global basis, the overall interaction in the session. Specifically,
dialog management determines the protocol of who talks when and how this control of
talking is passed from one user to another.
Activity Management: An activity can be defined as the totality of sequences of events that
may be within a session or may encompass several sessions. From the applications
perspective, the application can define a sequence of events called an activity and the
session service will ensure that it will monitor and report back if the activity is completed
or if it has been aborted that such is the fact.
For example, in a medical application, we can define an activity called "diagnosis" and it
may consist of a multiple set of session between several consulting physicians. We define
a beginning of the activity when the patient arrives for the first visit and the end when the
primary physician writes the diagnosis. The session service will be responsible for ensuring
that all patients have a "diagnosis".
Synchronization: We have seen that at the heart of a multimedia system is a multimedia

data object. Each of the objects has its own synchronization or timing requirements and
more importantly, a compound object has the orchestration requirement. The session
285 | P a g e
service of synchronization must then ensure that the end to end timing between users and
objects is maintained throughout.
Event Management: The monitoring of performance, isolation of problems, and restoration

of service is a key element of the session service. Full end to end network management
requires not only the management of transport and sub network, but requires that across all
seven OSI layers, that overall end to and management be maintained (See McGarty and
Ball).
Here we have shown the session entity which is effectively a session service server. The entity is
accesses from above by a Session_Service Access Point (S_SAP). The session entities
communicate through a Protocol Data Unit (PDU) that is passed along from location to location.
Logically the session server sits atop the transport server at each location.
The servers are conceptually at a level above the transport level. We typically view the transport
servers as communicating distributed processes that are locally resident in each of the transmitting
entities. This then begs the question as to where does one place the session servers. Are they local
and fully distributed, can they be centralized, and if so what is their relationship to the Transport
servers. Before answering these questions, let us first review how the session services are accessed
and how they are communicated.
Session services are accessed by the higher layer protocols by invoking session service primitives.
These primitives can invoke a dialog function such as Token_Give. The application may make the
call to the S_SAP and this request may be answered. There are typically four steps in such a
request, and these are listed in Stallings who shows that the requests are made of the session server
by entity one and are responded to by entity two. The model does no however say where the session
server is nor even if it is a single centralized server, a shared distributed server, or a fully distributed
server per entity design. We shall discuss some of the advantages of these architectural advantages
as we develop the synchronization service.
8.3 DIALOGUE MANAGEMENT
Dialog management concerns the control of the end user session interaction. Specifically, who has
permission to speak and when, who can pass the control and how is that implemented. In this
section we shall describe the environment for the dialogue management function and develop
several possible options for implementing this function.
Dialog management requires that each of the virtual users have a token or have access to a baton
in order to seize control of the session. In the course of a typical session, the two virtual users fist
establish the initial sub session that becomes the first part of the session. The addition or binding
of other virtual users through sub sessions to the session allows for the growth of the session. The
286 | P a g e
baton or token may be a visible entity that is handed from one to the other or it may be hidden in
the construct of the applications.
Consider the session level service called dialogue. The service can be implemented in four possible
schemes. These schemes are:
(1) Hierarchical: In this scheme there is a single leader to the session and the leader starts
as the creator of the session. The baton to control the session can be passed upon request
from one user to another, while full control remains with the session leader. The session
leader may relinquish control to another user upon request and only after the leader decides
to do so. The leader passes the baton from users to user based upon a first come first serve
basis. It is assumed that each users may issue a request to receive the baton, and that any
requests that clash in time are rejected and the user must retransmit. There transmit protocol
uses a random delay to reduce the probability of repeated clashing. The leader always
acknowledges the receipt of the request as well as a measure of the delay expected until
the baton is passed.
(2) Round Robin: In this scheme, the baton is passed sequentially from one user to another.
Each user may hold the baton for up to Tbat sec and then must pass the baton. When the
baton is held, this user controls the dialogue in the session.
(3) Priority: In this case, all of the users have a priority level defined as Pk(t), where k is
the user number and t is the time. We let the priority be;
Pk(t) = Rk(t) + Tk(t) + Dk(t)
Here R is the rank of the k the user, T is the time since the last transmission and D is the
data in the buffer. We assume that some appropriate normalization has occurred with this
measure.
Every Tcheck seconds, each users, in sequence sends out a small pulse to all other users,
on a broadcast basis, and tells them their current priority. Each user calculates the
difference between theirs and all the others. User k calculates a threshold number, TR k,
which is;
TRk = max |Pk(T)-Pj(T)|
If TRk > 0, then user k transmits its packets for Tsend seconds.
287 | P a g e
(4) Random Access: Each user has a control buffer that indicates who has control of the
session, namely who has the baton. The session is broken up into segment Tsess in length,
with Treq seconds being relegated to a baton ownership selection period and Tsess-Treq
being left for the session operation. During Treq , all of the users transmit a request packet
that is captured by all of the other users buffers. Treq is broken into two parts, Tsend
and Teval. These requests are broadcast in Tsend.
Now after the sent messages are received, one of two things can happen, the message is
received or it collides with another message and is garbled. If the message is garbled, the
buffer is not loaded and is left empty. If it is filled, then each buffer during Teval
sequentially broadcasts its contents and all of the users listen to the broadcast and record
the counts, Nk where Nk is the number of votes for user k in that call period.
The choice of baton control is then;
Choose user k if Nk = max j | Nj |
else restart Treq again.
For each of the protocols we describe the advantages and disadvantages of each in Table 4.3.
288 | P a g e
Table 4.3 Dialog Protocol Comparison
Protocol Advantage Disadvantage

Hierarchical Single Point of Control of the Lacks capability to have open
Session. discussion.
Priority Establishes who is in charge Requires a scheme to give

by allocation. priority that may be open to
compromise.
Round Robin Everyone gets to talk. May be excessively time

Egalitarian approach. consuming.
Random Strongest player wins. May not permit dissent.
8.4 ACTIVITY MANAGEMENT
Activity management looks at the session as an ongoing activity that users may come and go to.
This services provides an ability to easily add, delete and terminate the entire session.
An activity in the terms of the session is a total bounded event that can be compartmentalized in
such a way that other events may be locked in suspension until that event is complete. Activity
management is in the session layer a function similar to transaction management in a transaction
processing system. It allows for the definition of demarcation points that permit suspension of
activities in other areas until the activity managed transaction is complete. Activity management
can also be developed to manage a set of events that can be combined into a single compound
event.
There are several characteristics that are part of activity management:
Activity Definition: This allows for the defining of an activity as composed of several
dialogue. It allows for the defining of the activity as a key element of a single session or
even to expand over several sessions.
289 | P a g e
Activity definition is the process of informing the session server of the beginning and end
parts of an activity and in providing the session server with an identifiable name for the
activity.
Activity Integrity Management: Activities are integral elements of action that cannot be
segmented. The activity management system must ensure that once an activity is defined
and initiated, hat no other activity that could interfere with the existing one is allowed to
function.
Activity Isolation: The ability to provide integrity is one part of managing the activity.
Another is the ability to isolate the activity from all others in the session. An activity must
be uniquely separable from all other activities, and this separation in terms of all of its
elements must be maintained throughout its process.
Activity Destruction: All activities must be destroyed at some point. This is a standard
characterization.
There are several sets of activities that are definable in a multimedia environment. These are as
follows: Compound Multimedia Object Transfer, Sub-Session, Management, Dialog Control
The algorithms to perform the activity management functions are developable consistent with the
OSI standards. There are no significant special development necessary.
8.5 SYNCHRONIZATION MANAGEMENT
Synchronization is a session service that ensures that the overall temporal, spatial and logical
structure of multimedia objects are retained. In this case we have a source generating a set of Voice
(VO), video (VI), and Image (IM) data objects that are part of a session. These objects are simple
objects that combined together form a compound multimedia object. The object is part of an overall
application process that is communicating with other processes at other locations. These locations
are now to receive this compound object as show with the internal timing retained intact and the
absolute offset timing as shown for each of the other two users.
290 | P a g e
Figure: Synchronization
Network t 0 +T t 1 +T
t0 t1 k k
Voice
Video
Image
t 0 +T t 1 +T
m m
In this example, the synchronization function provided by the session server to the applications
processes at the separate locations is to ensure both the relative and absolute timing of the objects.
The location of the functionality can be centralized or distributed. Let us first see what the overall
timing problem is. Consider a simple SMO synchronization problem. The network than transmits
the packets and they arrive either in order or out of order at the second point. The session server
must then ensure that there is a mechanism for the proper reordering of the packets at the receiving
end of the transmission.
Let us consider what can happen in this simple example.
First, if the BMO of the SMO is very lengthy, then as we packetize the message, we must
reassemble it in sequence for presentation. Let us assume that the BMO is an image of 100
Mbits. Then let us assume that the packet network has a packet delay that will be low if
there is no traffic and grows as traffic increases. Now let us assume that the network
provides 500 bit packets transmitting at 50 Mbps.
Second, let us note that there are 200,000 packets necessary to transmit the data. Each
packet takes 10 microseconds to transmit. If we assume that there is a load delay of 5
microseconds per packet, then the total transmit time goes from 2 to 3 seconds.
291 | P a g e
We can also do the same with a compound object. In this case, we take the CMO and note that it
is composed of SMOs. The SMOs must then be time interleaved over the transmission path to
ensure their relative timing. It is the function of the session service to do this. The application
merely passes the CMO and its header information as a request to the session server to ensure the
relative timing is maintained.
The architecture for the session synchronization problem is shown. Here we have a CMO entering
the network, knowing that the session server at Server 1 must not only do the appropriate
interleaving but it must also communicate with the other servers (in this case K and N) to ensure
that de-interleaving is accomplished. We show the session servers communicating with the
network through the T_SAP and that in turn takes care of the packetizing. However, we also show
that the session server, 1 and N, communicate in an out of band fashion, using some inter process
communications (IPC) scheme, to ensure that the relative actions are all synchronized amongst
each other.
Figure: Synchronization Architecture
T_SAP
O C CMO SS 1
Network
O C CMO SS N
T_SAP
We can now envision how the architecture for this can be accomplished. There are two schemes:
Centralized: It assumes that each application (A) has a local client (CL). The application
communicates with the local client (CL) to request the session service. The session server is
centrally located and communicates with the application locally by means of a client at each
292 | P a g e
location. This is a fully configured client server architecture and can employ many existing
techniques for distributed processing (See Mullender or Couloris et al).
Figure: Centralized Architecture
A1 A2
CL 2
CL 1
Session
Requests Server
CL n
An
Distributed: In contrast to the centralized scheme, we can envision a fully distributed session server
architecture as shown. In this case we have a set of applications, and cluster several applications
per session server. We again user local clients to communicate between the session server and the
applications. The clients then provide local clusters of communications and the session servers
allow for faster response and better cost efficiency. However, we have introduced a demand for a
fully distributed environment for the session managers to work in a distributed operating system
environment. As a further extreme, we could eliminate the clients altogether by attaching a session
server per applications and allow for the distributed processing on a full scale.
Figure: Distributed Architecture
293 | P a g e
A 1,n
CL 1,n
A 2,1
CL 2,1
SS SS
SS22
SS11
SS
SSNN
CL k,N
A k,N
The major functions of the session server in its synch mode are:
1. To bind together simple objects into compound objects as requested by the application.
2. To provide intra object synchronization to ensure that all timing within each object is met.
3. To orchestrate amongst objects to provide inter object timing.
4. To minimize delay, slippage, between simple objects.
5. To minimize delay, latency, between different users.
To effect these requirements, we have developed and implemented a scheme that is based on a
paradigm of the phased locked loop found in communications (See McGarty and Treves,
McGarty). Here we have a distributed session server architecture receiving a CMO from an
application. The session server passes the message over several paths to multiple users. On a
reverse path, each server passes information on the relative and absolute timing of the CMO as it
is received using the session services primitives found in the OSI model. Generally for segmented
BMOs this is a simple problem but with streamed BMOs this becomes a real time synchronization
problem.
Figure: Synchronization Architecture
294 | P a g e
SMO
SMO
CMO SS
SS Transport
CMO Transport
CMO
CMO SS
SS
CMO
CMO SS
SS
Here we show M session servers and at the sending server we do the pacing of the packets to the
T_SAP and allow for the interleaving of the SMOs. Based on the commands from the feedback
system we provide delay adjustment, through caching and resetting priorities to the T_SAP for
quality of service adjustments for the lower layer protocols.
295 | P a g e
Figure: Detailed Synchronization Implementation
SS 2
CMO SS 1
Network
SYNCH SS M
SPC 2 SPC N Read Synch Pulse

Check SMO Timing
Check CMO Timing
At the receiving session servers, the synch pulses are read by the server, the SMO timing errors
are read, knowing the synch header, and an error message is generated. We also do the same for
the inter object CMO timing error.
The information is sent back in an out of band fashion to the source session server which in turn
controls the synch control pulses for the source session server.
We can provide further detail on the synchronization scheme as follows:
A CMO is generated by the applications program. This may be a totally new CMO or a result of a
new SMO addition or deletion.
The Source Session Server (SSS) transmits the header of the CMO to the Receiver Session Servers
(RSSs). They then respond with an acknowledgment and in turn set up their internal timing and
sequencing tables for local control. They also use the CMO header to adjust their local clock for
network timing references.
The SSS commences to interleave, sequence and pace the SMOs of the CMO down to the T_SAP
for transport across the network. At this point, the Transport protocol must have certain
requirements of either increasing bandwidth (e.g. local data rate requests and also controlling
296 | P a g e
sequence order. This interaction between the SSS and the T_SAP will define what additional
capabilities we will need at the Session layer.
At indicated instances, the SSS inserts local synch pulses in the interleaved CMO. The synch
pulses are to be used as local timing reference point for global coordination.
The RSSs read the local synch pulses and relates them to both the SMO and the CMO and obtain
offsets from the global system clock that has been updated in the RSS. It then send back the offset
of the synch pulses on a periodic basis. The offset is a vector that is given by:
E(k,j) = [e(k,j,1),.....,e(k,j,n),e(k,j,M)]
where E(k,j) is the offset vector of RSS j at time instant k. The internal values of the vector are the
offsets of each of the SMO elements and the last entry is the offset of the CMO.
The SSS uses the set of E(k,j) for j=1,..,N RSSs to calculate an overall error signal to control the
SSS. There are two major control features. If the average error is low then the SSS can reduce the
insertion of synch pulses and the lower the processing load. If the errors are large, then more synch
pulses are inserted to obtain finer loop control. The second element is control over the lower layers.
We use the magnitude of the delay offsets to send messages to the T_SAP to change the quality of
service parameters for the system.
We have developed several performance models for these protocols and the architecture that has
been developed to implement them.
8.6 EVENT MANAGEMENT
Event Management deals with the overall end to end management of the session. It is more
typically viewed as a higher level network management tool for multimedia communications. In
the current sate this service is merely a reporting mechanism. Although ISO has expanded the
network management functionality of the seven layers, most of the functionality is still that of
event reporting. In this section we discuss how that can be expended for the multimedia
environment.
Event management at the session layer provides for the in band signaling of the performance of
the various elements along the route in the session path as well as reporting on the status of the
session server and the session clients. We note that each entity in the session path, which is limited
to all involved clients and all involved servers provide in band information on the status of the
session. In particular the in band elements report on the following:
Queue size at each client and server. The queue size can be determined on an element by element
basis.
297 | P a g e
Element transit and waiting time. For each element involved in a session, the time it takes to transit
the entire block as well as the time that the block has been in transit can be provided.
Session synchronization errors can be reported in this data slot. These errors can be compared to
lower level errors and thus can be used as part of the overall network management schema.
The structure of the event management system has been effectively demonstrated. It is represented
as a header imbedded in the transit of every data block. We can generate specific event
management blocks that are also event driven and not data transit driven. These are generated by
direct transmission of such blocks as overhead devoid of data content.
8.7 CONCLUSIONS
What we have shown in this paper is that the session layer functions are key to supporting the
overall needs of a multimedia communications environment. We have also developed algorithmic
approaches for dialog and synchronization services and have shown that these services depend
upon the lower layers for support. Specifically, we have shown that if the underlying
communications network is jittery in the packet transport provided, the resulting delays associated
with the synchronization process can be significant.
Architecturally, we have raised several issues as to how best to provide the session service,
specifically where to place and how to communicate with a session server. The session services
require considerable entity to entity communications and this may require a distributed
environment of session servers all functioning in a fully distributed mode. In the network
applications developed to date (See McGarty and Treves), the session server has been centralized
and has allowed for communications in a distributed fashion on a UNIX environment using sockets
(Berkeley 4.3). However, in future implementations, the session server will be architects in a more
distributed fashion.
298 | P a g e
9 DISTRIBUTED DATABASES
In developing the multimedia communications network, there were many elements that were left
defined but not addressed. Specifically, we discussed many of the higher layer services that can
be provided in sessioning, presentation and application layers. In a multiuser environment, these
services must be implemented to support a fully distributed capability.
Whenever a network is working across several departmental, organizational and geographical

boundaries, it becomes inefficient to have a fully centralized system design. In those cases, it is
more appropriate to have a fully distributed design.
This chapter discusses the issues and alternatives of distributed elements. In particular we focus
on distributed data bases, distributed operating systems and distributed processor configurations.
We develop these in the overall context of the architectural requirements of such distributed
architectures. In addition, we focus on understanding the performance factors and their tradeoff
factors.
9.1 ENVIRONMENT FACTORS
The distributed environment includes the combination of the distributed users and the ability to
include many of the systems resources into the ongoing session. The session will have the
capability to access the processors and the databases and should
be able to do so in as real time fashion as possible. The evolution of distributed environments has
developed over the past twenty years. The earliest system was with the development of the
MULTICS operating environment that was developed at MIT for the use in a time sharing
computer system. It further evolved with the development of IBM's SNA architecture in the early
1970's. Yet in all of these systems there was still a hierarchical system that controlled all of the
systems resources.
A true distributed environment requires that all of the resources and the users can have an
arbitrary but definable relationship to all other users in the system. For example, the session
concept allows for the development of a complex environment wherein any user in the session
can play the lead of follower roles and any resources available in a session can be shared in any
fashion with any user or sets of users.
299 | P a g e
The environment entails all of the resources being at several locations and that there is now
single point of management and control. The distributed environment is distributed in five
dimensions that we shall discuss in this chapter:
Processors: This is a physical distribution of the systems resources. Typically we envision the
location of computing resources to be a one location and the users share that resource in some
fashion. Current computer architectures allow for local distribution of assets such as is found in
the VAX cluster
concept. Research efforts have allowed for the fuller distribution of computing resources through
the use of simple distributed operating systems. However, the fully distributed processor
environment is possible with the use of the techniques that are discussed in this section. Thus no
single processor is in total control and the processors share resources in an almost parallel
fashion. We can envision this as an adaptive parallel processing environment, wherein the
connection between the parallel processors is logical and not just physical.
Processes: Distributed processes are already a common ' capability in many systems. The
distributed process environment allows for the management of the process execution over several
processors at the same time. This may be of use in both real time distributed systems as well as
in fault tolerant transaction designs.
Database: The distributed database environment has evolved over the past several years and
there is now a body of technology that can support such operations. The distributed database is a
key element in a distributed system. It becomes the backbone of operating in an environment
where there is data generated by many users and this data has time sensitive nature.
Operating System and Management: The operating system environment is the overall
mechanism for controlling the multiple system resources. With the use of distributed processors
and the
need for coordinating the set of all system resources, there is a need to do so in a fully distributed
environment. The overall management of these resources is done through the operating system.
Communications: The essence of the communications environment is the session. As we have

discussed, the session is the access mode for enabling multiple users to share resources, data, and
process applications. The distributed communications environment must be built upon the
overall structure of the distributed processors and operating system.
We have depicted two possible architectures. The first is the heterogeneous environment, one
that is more common in systems operation. The heterogeneous environment shows the need to
300 | P a g e
share such different resource as memory, CPU, data files, image storage devices and I/O devices.
The specific application envisioned is that of the printing and publishing industry.
The second environment is the homogeneous environment. This is the truly distributed processor
environment wherein we can envision multiple processor locations of similar devices all sharing
in the overall processor loads. An example is the internetting of multiple work stations so as to
allow the utilization of all the resources simultaneously.
Recall that a process is a program in execution. A process is typically associated with a single
processor and the functioning a process across multiple processors is generally not expected.
However, in a fully distributed environment, there is a clear need for such functioning. Consider
the following example.
In a medical application are, there is the need for radiological consults between the primary
physician, the neurosurgeon, the neuroradiologist and the oncologist for a patient with a primary
brain tumor. The consult is conducted in the context of a session and uses the resources of
several processors as well as several database. There is an application program that can take the
multiple radiologic data, MRI, CAT and nuclear scans, and combine them into a three
dimensional image of the patient. The neurosurgeon needs to manipulate this for the selection of
laser surgery procedure and the oncologist is interested for the impacts of blood brain barrier
effects for post-operative chemotherapy.
The process that is in operation is resident on each of the specialist processors and is shared
amounts them.
In this particular case we can envision the application being in the area of financial transactions.
The system shown is used for the checking of credit card validation and assuring that the seller
can obtain payment from the buyer. The system must track the activity of all the purchasers and
seller in the national network. In particular ,care must be taken to track the performance of any
single card and to see if it has been used an excessive number of times.
In present day credit card systems, the data base is centralized and allocate checking is done in a
hierarchical manner. The data base is accessed via a communications network and the overall
cost per transaction can be quite high due to the communications overhead. The distributed
database design allows for minimum communications and a distributed set of datafiles with full
interconnectivity between them. This system requires that a single card users has their activity
tracked as they move geographically through the system. One way to do this is to pass the file off
to the local database and then to act as if the system were a fully segmented design. The second
alternative is to keep parts of records at a distributed set of locations and to have a set of
composite numbers, such a total exposure per card tallied at all or a few locations.
301 | P a g e
We shall develop this concept in detail in latter sections. However, the overall requirement for a
distributed operating system is to allow for the overall management of the systems resources by
managing processes, managing data, memory and controlling events. Also the operating system
must allow foe control of the overall I/O resources of the system.
The communications issues in a distributed environment revolve around the need to provide a
seamless connection between users in such a way as to make the presence of the communications
network appear to result in minimal system impact.
Communications in a distributed environment has evolved over the years. As we have shown in ,
there are standard protocols to handle this communications issue, hate we wish to show in this
chapter is that the interaction of database, operating systems , processors and processes may
require a modification of such standard communication architectures. In particular the time
required of all the protocol overhead may be prohibitive in the operation of a fully distributed
system.
We can combine these separate elements and show how they interrelate. The specific example
that we have used is that of a Hospital information management system for the care of patient
records and one that connects multiple teaching hospitals into a single network.
9.1.1 Network Requirements
The network requirements focus on the needs for the network to, provide more than just a data
path. The requirements are driven by the high data rates and bandwidth available to the end user
and the ability to configure the channel in a more fluid fashion.
9.1.2 Session Requirements
The session was the key element of the multimedia multiuser communication -paradigm.
Sessions have been discussed in detail in their many element. In this section we shall develop the
overall requirements on the session concept when it is impeded within a fully distributed system.
Hierarchical schema for part of the COMPANY database have level zero. The level of a nonroot
node is one more than the level of its parent node. A descendent D of a node N is a node connected
to N via one or more arcs such that the level of D is greater than the level of N. A node N anil all its
descendent nodes form a subtree of node N. An occurrence tree can now be defined as the subtree
of a record whose type is of the root record type.
302 | P a g e
The root of an occurrence tree is a single record occurrence of the root record type. There can be a
varying number of occurrences of each nonroot record type, and each such occurrence must have a
parent record in the occurrence tree; that is, each such occurrence must participate in a PCR
occurrence. Notice that each nonroot node, together with all its descendent ra)des, form a subtree,
which, taken alone, satisfies the structure that the integrity constraints specified on the relational
database schema are not violate)!. In this section we discuss the types of constraints that may be
violated by each update operation and the types of actions that may be taken in case an update
causes a violation. We use the database shown and discuss only key constraints, integrity
constraints, and the referential integrity constraints for each type of update we give some example
operations and discuss any constraints that each operation may violate.
For the COMPANY schema if an interface between PASCAL and the networKpBMS were
available, it could create the PASCAL program variables. A single record of each record type can
be copied from or written into the database usinji the corresponding program variable of the
UWA. To read a record from the database, we useJthe GET command to copy a record into the
corresponding prograrrm variable. Then we can refer to the field values to print or to use for
calculations. To write a record into the database, we first assign its field values to the fields of the
program variable and then use the STORE command to actually store the record in the database.
Currency Indicators
In the network DML, retrievals and updates are handled by moving or navigating through the
database records, and hence keeping a trace of the search is critical. Currency indicators are a
means of keeping track of the most recently accessed records and set occurrences by the DBMS.
They play the role ofsposition holders so that we may process new records starting from the ones
most recently accessed until we retrieve all the records
9.2 DATABASES
Databases are key elements for the storage of the multimedia data elements. The database issue
has received considerable analysis and development for the standard computer files that exist
today. These are typically local data bases that are accessed either locally or remotely. The key
issue of multimedia communications is the need for both distributed database as well as
multimedia databases. The issue of a distributed database is expanded now that we have the
capability to transmit data at Giga bit rates. We can now reduce the access time per file for
transport to a factor less than that for disk access. This opens new issues for the design of such
databases. The multimedia data base represents a new and innovative method of storing
multimedia storage. It has two dimensions. The first is the fact that the data may be stored on
different media themselves and that may require a nonhomogeneous access and storage protocol.
The second is that a multimedia element may no longer be bounded as simply as we have known
such computer files as bank records of income tax records. Thus the multimedia database issue is
303 | P a g e
one of inhomogeneity and unboundedness. Both of these characteristics are typically lacking in
standard data bases.
9.2.1 Database Structures
Databases are typically structured collections of data elements that are to be processed on an as
needed basis by the end user. Quite simply, we may think of a database as a collection of
elements that may be capable of representing the totality of information needed for a specific
application. For example, consider the data base that may be needed for the distribution of
Christmas cards. Such a data base may need the name, address (street, city, town, and zip code),
the relationship ( whose side of the family and the specific relationship, such as aunt), the status
of sent or received from in the past year, and the gifts that may have been sent or received. What
we see in this simple example is that each element of the data base is definable and also is
definable in terms of a certain well defined alphanumeric quantity.
In a multimedia database however, we do not have as simple a set of structures. Consider a

similar example but for a multimedia case. In this example we are trying to track the progress of
a patients recovery from a head trauma sustained in an accident. The characterization is that of
patient John Doe. The data associated with this patient includes, X-rays, MRI scans, CAT scans,
oral records of the radiologist and neurologist, written records from the primary attending
physician, and the cardiologists electrocardiograms. What we immediately see is that the records
are not so readily characterized as a finite set of alphanumeric data. The data are video, voice,
text, graphics and
image data. The data in addition are of significant and varied length, and the relationship
between the different media are suggested but not precisely structured.
9.2.1.1 Data Base Elements
A standard structure for database is the entity-relationship (ER) model. An entity is any real
object that one desires to discuss and store data on. For example, the patient may be en entity.
Associated with any entity is a set of attributes. In the patient example, the attributes may be the
name of the patient, the address, age , diseases, and attending physician. We can see that any
attribute may be further extended as its own entity with its own attributes. Thus in the present
case, we have the disease and it may be characterized by organ, name, level of involvement.
Attribute are generally of bonded form, that is the patients name, has a first, middle and last part,
and each part is composed of a set of ASCII characters that do not exceed 50 in length. Thus the
name can be at most 150 characters or 1200 bits.
304 | P a g e
Consider now a multimedia data base in the medical application area. The entity is the patients
radiology report. The patient had a set of X-rays and an MRI scan. The "radiology report" entity
has the following attributes:
X-rays, including 5 scanned images
MRI images, three sets of 12
Written report of radiologist
Written report of neurologist
Written report of pathologist
Oral report of radiological consult
Oral report of attending physician
In the case of a standard data base, the attributes have a finite set of values that they can take.
These finite set of values are called the value sets or domains of the attributes. Names can be up
to 50 characters for example. However, the domain of an attribute such as the oral report from a
radiologist may be less well-structured and in fact may be unbounded.
We can envision that there can be multiple entities in a large data base. Again consider the
hospital data bases that we have been developing. Consider several entities: Patient, Physician,
Insurer, Procedures, Prescriptions
The patient entity may contain the attributes: Patient Name, Address, Age, Insurer, Physician ,
Phone
The physician entity may contain the following attributes: Physician, Name, Address, Phone,
Specialty, Backup Physician, Specialty
We can now relate these two entities through a RELATIONSHIP called Patient_Of. Thus we
have:
PATIENT Patient_Of PHYSICIAN
This shows that there are two entities that are related through the relationship. This types of
relationship is called a binary relationship because it relates one entity type to one other entity
305 | P a g e
type. There are relationships, that is action statements that relate one entity to another, that ca
relate multiple entity types. Those that relate three entity types are called ternary etc.
Consider now a set of three multimedia entities that are prepared for a patient. They are:
RADIOLOGY REPORT
PATHOLOGY REPORT
CARDIOLOGY REPORT
Each of these reports contain attributes the are text, image and voice. We can also have
standard entities that can be used in a normal data base. These entities are:
PATIENT
PHYSICIAN
INSURER Consider now two relationships:
Requests_Report
Prepares_Report
We can now have the following statements:
PATIENT Requests_Report RADIOLOGY REPORT
PHYSICIAN Requests_Report RADIOLOGY REPORT
INSURER Requests_Report RADIOLOGY REPORT
Thus we can see that data base are merely ways of storing recorded facts, called attributes, in
well-defined bundles called entities, and associating them with each other in relationship. The
structure of the language that uses the data base is called the data base language and has its own
syntax. However, the syntax is governed by the underling structure of the database.
9.2.1.2 Relational Databases
306 | P a g e
A relational database is a means of characterizing the data and structuring it so that it may be
accessed and altered in a fashion that allows for maximum flexibility. In a relational database,
we develop a flexible set of relations and build these from the bottom up to have as flexible and
accessible format as possible.
The basic building blocks of a relational database are as follows:
Domain of Atomic Values: These are irreducible entries that represent final data values, such as
the age of a patient, their first name, or their phone number.
Attribute: This is the name or role played by a particular domain. Thus phone number is the
attribute name played by the domain xxx-xxx-xxxx.
Relation Schema: This is a set of attributes, {Al,...,An} that has a name R, which may be the
term patient, and has a degree equal to the number of attributes, or irreducible elements.
Relation: A relation is a set of tuples, where a single tuple is the set of attributes of the relation.
For example, the relation schema, address may contain, the attributes, number, street, city, state,
and zip. This is a 5-tuple. The relation ADDRESS now is the set of all such tuples that are
relevant to this application. It could be zero, one or even one million addresses.
Tuple: As we have just noted this is a specific relation entry. In the above case for address, we
have:
t = {24, Wood St, Greenwich, CT, 08534} or for the general case:
t = {number, street, city, state, zip} and the relation:
ADDRESS = {tl,t2,,tn}
Key: The key of a relation is an attribute that can be used to uniquely identify the relation. In the
case of address, we may use the zip as that attribute if and on if we know that people are in one
town at a time. There may be many keys in relationships, and there can always be a primary key.
We can now see how relations can be referred to one another. Let us define three relations
through the three schema:
R1(A1,A2,A3,A4,A5)
and rl = {tl,t2,t3,t4,t5,t6,t7,t8,t9}
307 | P a g e
where ti are all 5-tuples o R2(A6,A7,A1)
and r2={ul,u2,u3,u4}
where ui are all 3-tuples and Al is the same as in rl.
R3(A8,A5)
and r3={vl,v2,v3,v4,v5,... ,vll4}
where vi are 2-tuples and A5 is the same as rl.
We can now point from r3 to rl and from r2 to rl. This pointing is through the common attributes
in both cases.
There is an algebra that can be developed using the relations that entails the use and definition of
operations on the relational schema. These operations are of the type: Union, Difference, Select,
Divide, and many others (see Elmasri and Navathe)
9.2.1.3 Hierarchical Databases
The hierarchical data structure is an older scheme of database and follows the need for a
structured environment. In the relational case we define general relations, many of which can
change and evolve in time and are related to one another through keys. In a hierarchical scheme,
we have a much more structured system that is related a priori.
In the hierarchical scheme we have several elements that make up the data base paradigm. These
elements are:
Record: This is the collection of data that are the basis of the database. The record could be the
patients account and would be composed of name, address, age, etc.
Parent-Child Relationship: This is the relationship between one record and other records. For
example, we can say that Department (eg Cardiology, Radiology, Urology etc) cane be the
highest record in the hospital. This may relate down to Patient and Physician. In turn, Patient
relates down to Tests' and Insurer records.
Field Type: This is the detailed contents of the Record type. In the patient Example we may
have:
308 | P a g e
Patient = {Name, Address, Age, Disorder, SS Number}
The patient Tests record may include the following Field Types:
Test = {Type, Date, Physician, Result, Next Test Date}
Hierarchical schemes have a fast access time for most applications but have many major
drawbacks. It is generally difficult to change data elements or records and we have to be cautious
in handling non 1:N relationships in the parent child area. In addition, as we have noted in this
example, the patients are grouped by department first and to get statistics on patients, we must
sum across departments, not across patients.
The patients, and tests etc are segmented by the hierarchical scheme.
9.2.1.4 Network Databases
The network database structure owes its early existence to Bachman and it represents a way to
structure the relationships between data records in a logical fashion on the overall data structure.
In the hierarchical scheme we had developed the notion of an overlaying hierarchical
relationship between all elements of the database. In the relational scheme, we had much greater
freedom in defining data elements and then after the fact overlaying relationships in a much freer
manner. In the network scheme, we find that there is a middle ground in which we do not have
the hierarchical structure but that there is an overriding relational structure in the database that is
called the networking of the data elements.
The network data base is composed of the following elemental concepts:
Data Values: These are the actual numerical or letter values that are entered into the data base.
Example would be specific names or numbers.
Data Item: This is the name ascribed to the data value. For example, we can use NAME, or
ADDRESS.
Record: This is a collection of data values in an organized form. It may represent all of the data
values stored on a particular patient.
Record Types: This is the name used to identify a collection of records, such as PATIENT or
PHYSICIAN.
We can then find a way to relate these together through the following concepts:
309 | P a g e
Set Type: A relation between two record types. One of the record types is called the owner and
the other is the member. There is only one owner and there may be several members. A set type
may be the one Treated_By and it relates PATIENTS to PHYSICIANS.
9.2.2 Data Base Access
The structure of databases is only one of the key factors in understanding the usage and
implementation of a database. The second element is the access to the data records themselves.
In this area there are three major issues of concern. The first is the issue of the physical access to
the data. Specifically what medium is the data stored on and how is the data physically stored on
that medium. As with the file structure, we consider the more common forms of file structures in
this section and defer the issues of multimedia databases to latter sections.
The primary form of data storage is the magnetic disk for linger term storage and random access
memory for shorter-term storage. The issue of medium is significantly impacted upon by the cost
of random access memory and its ability to be supported is a nonvolatile environment. With the
advent of multi-megabyte chips and the dramatic cost reduction in cost per bit, there are many
tradeoffs that can be made in the selection of physical storage medium strategy. We shall not
discuss those elements in this text but refer their reader to XXX.
The second issue that we shall discuss in this section is the issue of the logical access to
databases. Specifically, what languages do we use to readily access the information. The
languages typically used here are called query language and typical example are those of SQL
and the other languages that have developed syntaxes for access of structured databases. The
query language is the user interface issue that we have discussed at length in other chapters. The
query language has develop for structured database from a set of typically arcane commands to a
set of commands that today resemble a natural language. It is actually possible to ask such a
question as "How many people are there in the Drafting Department". A system may take this
sentence and parse it and then know enough to be able to recognize that it must add together all
the name files for a specific department.
Physical Access
Logical Access: Access Languages (DBL)
Database Access Interfaces
9.2.3 Distributed Databases
Distributed data bases are essential in environments where there would normally be significant
communications from one location to a distant one. A typical example that we have used is the
310 | P a g e
need for distributed database in the financial transaction processing area. Here we have to keep
track of many transactions that occur at various locations.
Let us first consider an example of such a distributed system of databases that is slightly
different. This example is that for the storage of 800 numbers in the telephone system. In this
example we will first deal with a passive database that is read by the user and there is no
transaction made in the database.
The 800 number system in the telephone network works on the principle that local users can dial
800-xxx-xxxx and can get a toll free phone call. The normal billing system will recognize a
prefix other than 800 as one to bill and will bill the calling party directly. When the 800 number
is referred to in a call, the number is looked at in a database table and it is cross referenced to an
actual area code number and that number is fed into the network as the dialed number. At that
point the end user does not know that he has skipped around the billing computer and has had his
dialed number converted. The conversion table could be placed at multiple locations or it could
be centralized. In this case the database is static. The difference between having this data base
centralized or distributed is the issue of call set up time, that is the time to place the call.
Let us examine the 800 database tradeoff. Consider the two cases of centralized versus
distributed. In the centralized case we have a large data base that is at one location. Let us
assume that there are 1 million calls per busy hour and that each 1 thousand calls requires a 1
MIP processor. The call set up time for this system is given by:
T SU = T TR + T I/O + T pro + T Resp + T TR
Here T TR is the transport time, T I/O is the I/O time and T resp is the CPU response time. For a
heavily loaded machine we have decreased I/O and response time.
In the 800 distribute case we assume that we can place an 800 processor at each of the 1,000
central office locations and this yields 1,000 MIP machines and reduced I/O per machine and
reduced processing time. In addition we have to split the database in such a way that we do not
have to multiply it 1,000 times. We do this by recognizing that only x% of the 800 numbers are
used by y% of the people in each of the locations. This is often called the 80/20 rule. Thus we
can save only a small percent of the conversions for translation and leave the rest for a
centralized system or route them to other systems.
Let us now consider a second example that is both read and written into in a distributed fashion.
This data base is one used for the buying and selling of commodities such as gold, oil, silver and
other such items. This differs from the first case because now there is a dynamic bidding process
that must
311 | P a g e
carefully take into account the timing of the transactions and assure the end user that bid and
asked prices for commodities are met.
Let us examine the process in some detail. There are sellers of commodities that present an
asking price. For example Seller 1 could say that they are willing to sell gold at $450 per ounce.
There are several buyers who are interested in the gold. However they are not willing to pay the
asking price. One bids at $448 per ounce as the highest. Then the seller reduces the price to
$449. One of the buyers sees this as an opportunity and bids • $446.50. The other buyer seeing
that this is a good deal bid $449,125 and the seller closes the deal.
Now there any be several markets for this gold. The markets may be in New York, London,
Tokyo, and Hong Kong. One approach is to have the data base used for recording and display the
bids to be centrally located in a city such as Chicago. Or alternatively, each city may have its
own data base and the system communicates amounts itself in a fully distributed fashion. Again
the tradeoff is time to transact and time to respond. In this type of market, the quicker I can
respond the better closure on the bid/asked price one can obtain.
A third example of a distributed database is that which can be used for the establishment of a
multimedia communications system, where sessions are established and users are sending images
and other multimedia objects from one point to another.
In this case, the system is used to set up calls and to maintain the multimedia sessions. The
session set up algorithm identifies all of the participants in the session and knows their locations
on the network. When a member of the session transmits a packet to a local node, the local node
accesses the overall database which knows the status of all participants in the session. It then
generates a set of additional packets, customized as is necessary and transmits them to the
desired nodes. It must assure that there is certain synchronization in the packet transfer and that
all of the packets are synchronously transmitted both interpacket and interuser on a single packet.
In these three example there are certain characteristics that arise that a distributed database must
reflect. These characteristics are:
Reduced access time due to the combination of reduced communications time and reduced I/O
time due to more efficient loading per processor.
Real time updating of data records and assurance of timing factors so that bid and asked numbers
are not reversed in time.
312 | P a g e
Overall synchronicity within and between packet transmissions to assure end to end
synchronicity for display and processing purposes.
9.2.4 Distributed Database Issues
We can now begin to discuss some of the structural elements of a distributed database. We shall
use the relational database system for the representation of database in this environment.
Approaches using the other schema are also possible and are left to be developed in the
problems. The discussion in the section is based upon the work by Ceri and Pelagatti.
Recall that a relation, R, is a table that store data. The data is stored in columns called attributes,
Ai. The number of entries in the table of the relation is called the tuples in the data entry.
In the normal non-distributed environment, we have a collection of relations that generate the
overall database. The global schema is this collection of data as if we had no distributed
environment. Now we ask ourselves, how do we split this up into several pieces so that we may
locate the data in multiple locations. We do so in two steps:
Fragment: In this step we take the global schema and generate separate disjunct schemas call
fragments and named Rli••••fRri•
Allocate: We assume that the database is to be distributed to k locations, Ll,...,Lk. We then must
map the fragments onto these locations. Many fragments may be mapped onto several locations.
We can deal now with the first issue of the allocation of the databases to many locations and
focus first on the issue of reading of such databases. Consider the database composed of many
sub elements. We desire to distribute this database over many locations. Let us take the example
of the 800 number database that is used in the telephone business. This database is a conversion
data base that converts the 800 number into a real dialable number of the type xxx-yyy-zzzz.
Rather than having this database at one central location and having all customers dial that
location, we can break this database apart and distribute it at many locations. The breaking apart
function is the fragmentation task, and the placing of these parts at multiple locations is the
allocation task.
In fragmenting a database, we are considering the ey need to break it apart into smaller segments
that can be accessible in some particular fashion. In the 800 number case we may segment by
other areas of the country, by the specific application, by the type of business or other reason.
When we fragment, we than may want to assign or allocate the fragments to multiple locations.
Thus if we define D as the total database, we can define D as:
313 | P a g e
D = {D1,D2,....,Dn}
where Dk is a specific fragment. We can then define Li as the ith location. Let Si be the
assignment vector for the ith location;
Si = {Sil,Si2,,Sin}
where :
1 if Dk is at Li
Sik =
0 otherwise
Then we have the database at Li being the following: D(Li) = DT Si
where T is the transpose operator, and we represent it as:
D(Li) = {SilDl,, SinDn}
The optimal segmentation and allocation problem can now be stated as follows. How do we
choose the best fragmenting and allocation scheme to either maximize or minimize some specific
performance criteria. One approach is to define the cost of a scheme and then to find the
fragmentation and allocation that minimizes the cost. Consider the following cost equation. Let
CTOT be the total cost of the scheme. The cost equals:
CTOT = CDB + CPRO + CCOM + CLAB
where:
CDB is the cost of the actual storage databases at each location and duplicated as needed at each
location.
CPRO is the cost of the processing power needed at each location to manage the database entry
and control.
CCOM is the communications cost of the system as needed for the users to access each of the
fragmented databases.
314 | P a g e
CLAB is the labor cost per access, summed over all access. It is a measure of the response time
of the system in providing a unit access to a datafield.
Cost for Memory: As we fragment the database, we then allocate it to many locations. The cost
for memory is dependent on how much we place at each location. At one extreme, we can
fragment and place only parts at one and only one location. At the other extreme we can
fragment and place all parts at all locations. The latter will be an upperbound on memory costs
and the former a lower bound. Depending on the cost of memory the cost may be significant.
The memory cost can be made lower by a slower memory storage system, however, that will
increase the labor cost element.
Cost for Processing: As we have done for memory, the processing cost is dependent upon two
factors, the number of accesses per location and the amount of memory controlled per location.
If we have few locations then we are dominated by access, if we have many locations but
significant duplication of the database, then we have significant access processing.
Cost for Communications: This factor depends on how many locations there are and how many
locations have the data base elements. If there are many locations and many data base elements
then the communications costs are low. If there is only one location then we would expect high
communications cost.
Cost of Labor: The labor costs are based on how many people are required for a volume of work.
The people are determined by the number of units and the holding time per person. If the overall
processing time is dominated by the response time of the human then this is the lower bound. If,
however, the response time is dominated by the system delays, then this is a factor. For example,
if we have log communications delays, long access times etc, we have large labor costs.
Conceptually, we can conceive of an optimization criteria that says the if this cost is to be
minimized, we could then determine the optimum fragmentation and allocation to reach the
minimum cost element. Generally this is not readily determined and suboptimum approaches are
performed. We demonstrate several of these in the problems.
We can now consider the simple second issue that relates to the real time updating of databases.
This is the issue of writing or reading and writing databases. In the previous analysis we have
developed an approach to the selection of how to distribute the database based upon the reading
of the files. In this area we shall focus on the issues of actively changing the contents of the
database. There are two major issues that are key to understanding the active change of a
distributed database. The first issue is that of recovery from faults and the second is the
concurrency of a database use. Recovery relates to the issue of how to deal with system errors
and how to restart the database after such a failure. The issue of concurrency relates to the need
315 | P a g e
to establish the concept of a transaction, which we shall do shortly, and show that transactions
are key elements in read write actions.
The issue of recovery of the database after a failure must be considered. The distributed database
is composed of three levels of elements. At the highest level is the data agents that control the
flow of the actions of the database. The agents have a define root agent that provides an overall
centralized control element. At the second level is the distributed transaction monitors DTM that
provide for the communications amounts the different locations. The DTM provide control of the
elements in the third level, the Local Transaction managers, LTM. The DTM elements
communicate amounts each other and also sent the action signals to the LTM.
The recovery issue relates to a transaction that is occurring that may be halted in mid-stream. If
this is the case, it is necessary to remember that it was not completed and possible, but not
necessarily, where it left off. It is the function of the root agent, in conjunction with the TMs to
keep logs of the status of the transactions. The complete transaction has to be defined and the
progress of the transaction is followed through a log process, If a fault occurs, then the log is
used as a means to determine if the system is recoverable.
Concurrency is the issue that relates to ensuring that one event does not occur before another that
id dependent upon the first. We would not want to post money to an account before we checked
the validity of the depositing check, and if posted allow the customer to withdraw the amount not
yet checked.
The issue of concurrency control is also critical. As we had done with recovery, concurrency
depends on the three layer architecture. Now however, we need to have the agents talk with one
another and with the dTMws. We perform the concurrency control through a locking
mechanism, combined with a structured and time tagged transaction sequence. The transaction is
a complex set of reads and writes, that must be defined a priori. By tagging the transaction
sequence as a whole, and providing locking mechanism throughout the database schema, we can
assure that improper reads and writes do not occur. The cost may occur in the efficiency of the
database, access time.
We can now define a transaction as an atomic unit of access to the database. This transaction is
complete by completing all of the elements of the access or otherwise it is not complete. We can
define Tj as a transaction. We further define a transaction as a combination of reads and writes to
a data base. Let W denote the write function or operation and R the read operation. Let Dk
represent the data item on which the operation is to be made. We call Dk a data item and say that
Dk is part of relation Rj. A data item is that set of a relation that is materially affected by an
operation of a read or write.
316 | P a g e
We can now define a transaction as an ordered tuple of W and R operations as follows:
Tj = { Rj (Dk) ,Wj (Dm) ,,Rj(Dp)}
That is Tj is a collection of ordered reads and writes, at locations and on data elements k,m,...,p.
Transactions have four basic characteristics that must be observed:
Atomicity: A transaction is the totality of all of its parts and cannot be broken down into sub
elements.
Durability: Any system that performs a transaction must assure the operators that the results of
the transaction endure.
Serializability: When there are several transactions executing simultaneously, them the result
must be the same as if the transactions occurred in a hourly serial manner. That is if there are
three transactions TI, T2, and T3, then if we have:
TI = [W1,W1,W1,R1,R1]
T2 = [W2,W2,R2,R2]
T3 = [W3,R3,R3]
We can now define a SCHEDULE, S, as a sequence of W/R commands and one serial schedule
is the one;
S = {T1,T2,T3}
Another schedule is a sequence of mixed versions of Wi and Rk. Any admissible sequence form
a schedule must be serializable, or equivalent to the first S.
Isolatable: This means that incomplete transactions must not be effected or cannot effect other
transactions.
We have seen that we ensure the necessary elements of a transaction are preserved by ensuring
that the use of locks, time tags and the other elements that we discussed are used in a distributed
database environment.
In the system at any one time there are many transactions that
317 | P a g e
are occurring. That is there are Ti, Tj, , Tq. This clearly
shows the problem of the distributed database. The issues are as follows:
Concurrency: This relates to changing data elements at different places at different times, when
there is a need to assure that a sequence of changes be made in certain order. For example Ti
may have a sequence or ordered reads and writes on a set of data elements. Tk may have another
set of reads and writes. However there may be an operation in Tj that may proceed an operation
in Tk that will alter the way Tk is carried out. A simple example will give the case.
Let us assume that we wish to reserve a seat on an airflight. However another person at another
city wants to reserve the seat on the same flight. Let us assume that you get to the counter in
New York one minute before the contending person does in San Francisco. Hopefully, the
system is first come first serve and you get that last seat.
Let us first examine the transaction that goes into reserving the seat.
A read is made into the database to determine if a seat is available.
A write is made into the database to reserve the seat on the plane.
A read is made into the database to check the credit card, o A write is made into the database to
charge the card, o A write is made to a data base to issue the ticket.
Thus the transaction is:
TI = {R1(D1),W1(D1),R2(D2),W2(D2),W1(D3)}
Here we have Wl being a read at location 1 which is New York, and W2 being Phoenix. The
second transaction is:
T2 = {R5(D1),W5(D1),R2(D2),W2(D2),W5(D3)}
Now these transactions are occurring almost simultaneously but the reservation system is in New
York (1) and San Francisco (5). Let us see what could happen in this case.
1. New York checks and shows available.
2. San Francisco checks and shows available.
318 | P a g e
3. San Francisco books seat.
4. New York books seat.
5. New York checks credit
6. San Francisco checks credit
7. New York charges credit
8. San Francisco charges credit
9. New York gets default on ticket print.
10. San Francisco gets ticket
11. You get very mad !!!!!
Even though the data bases update instantly, we still have not kept the integrity of the overall
transaction. Admittedly this was a bad way to create a transaction, it however, is typical of what
can happen in large scale transaction systems.
The solution of the concurrency problem is time stamping the total transaction and ensuring that
all data node controller have the ability to update each other simultaneously.
9.2.5 Multimedia Databases
The multimedia database consist of several media that contain a complex interconnection of
video, voice data, text, image and other forms of information. As we discussed in the context of
the reactions database, there can be significant connections between many of the different files.
For example, consider a voice file that contains certain information on a particular patient record.
This file may be associated with a video file of the patients echo cardiogram. It is necessary to
ensure that the files are related to one another, that they can be tracked down and also
synchronized in time and space. These types of requirements are not typical for alphanumeric
records and files.
9.2.6 Physical Access
319 | P a g e
Let us consider the case of accessing the record of patients in a hospital. There are four types of
record that may be of concern for the development of a certain type of procedure. These records
are:
Patient records stored in the Relational database disks.
Radiological images stored in a CD juke box storage system and managed by a PACS type data
manager.
Text records stored in an image based CD system similar to the image files and front ended by a
separate database manager.
A voice storage system that is the storage vehicle for voice messages that annotate the image
files.
A video angiography of the patient's hear stored on a CD rom disk.
We can consider the multimedia data object as a meta-object which is the concatenation of the
set of four separate data objects. Let us define these objects as follows:
OR is the data object associated with the patients files
01 is the data object associated with the patients radiological records. Note that this may be just
x-rays, all x-rays, x-rays and MRI data, or a concatenation of many of these elements.
OT is the collection of text records that relate to this file. It may include the write ups on the
current x-rays prepared by the attending radiologist or any other set of elements.
OV is the voice samples of the consulting radiologist or other related physician.
We can now defining the compound data object associate with the patient as OP and definite it
as:
OP = {OR,OI,OT,OV>
where {} stands for an appropriate concatenation. We shall
discuss the structure of this concatenation in detail latter in this section.
320 | P a g e
We can now see how we can relate these compound data objects the physical system design by
examining the overall system architecture.
There are separate storage elements and assume for this case that they are generally located in
close proximity. There are two extreme strategies for developing physical access to these
records. The first assumes that each is a standalone system and the end user must establish all
connections to related records.
9.2.6.1 Logical Access
The logical access issues for multimedia database are significantly more complex than for the
alphanumeric databases. The SQL type languages for physical database are based upon the
simpler structure of the records and files and make the use of simple syntax possible. Queries in
a multimedia context are much more complicated. In an alphanumeric system it is possible to ask
the question "How many patients in 1989 have had acute endocarditis?". The statement is parsed
into the following elements:
How many: Count the number and the answer is an integer.
Patients: Count on the patient records
1989: Condition on the year
endocarditis: Condition on the disease.
Consider now the type of question we may query the data base with in a multimedia format.
"Which patients have shown a cholecystitis diagnosis that the physician has expressed concern in
the extent of the diagnosis and also the scans are not fully determinate?".
Let us examine this query and compare its elements to the query for an alphanumeric data base
query.
Distributed Multimedia Databases
MMDB Design Factors and Issues
9.3 CONCLUSIONS
This chapter developed the concepts of databases and further developed concepts for both
distributed database and multimedia databases. The database element is key to the storage and
retrieval of multimedia files. Knowing the standard data elements in a typical record file is
321 | P a g e
necessary but not sufficient for the development of a multimedia database. In we had developed
an understanding of the types of files that we could develop in a multimedia environment and
studied to types of storage media available. That allowed us to develop the times statistics
necessary for the operations of a data file retrieval. In this chapter, we have extended that to the
database element, understanding the way the information can be stored and what languages can
be used in its retrieval.
322 | P a g e
10 DISTRIBUTED OPERATING SYSTEMS
The control of the flow of information, requests and the overall management of the session
process is generally beyond the control of the typical functions of the communications systems
that we had developed in. This management can be viewed as a major function of the operating
system of a standard operating system of a stand-alone computer and more importantly the
function of a distributed operating system of a fully distributed environment.
There are two extreme approaches that can be taken to viewing the overall operational control
and management of a large distributed system. The first approach states that the control can be
fully distributed in the layer architectures of the separate elements of the network, and thus
follow all of the ISO layered commands. Frequently this works well within a purely record based
and generally homogeneous network of multimedia elements. The second approach states that a
form of central control, although implemented in a distributed fashion is a means to the same
end. This latter approach allows for the integration of multiple system elements, even though
they may not be fully ISO compatible. Moreover, it allows for the introduction of new and
innovative multimedia interfaces that the ISO level may not anticipate.
In this chapter, we focus on the development of a multimedia distributed operating system that
ensures that the sessions are managed properly and that handles all of the other functions of an
operating system. These functions include such elements as file and record management, process
and session management, device and I/O management and an overall network management
function.
As with the other chapters, our focus is to build on existing paradigms and then to expand the
concepts to a fully distributed and multimedia environment. Moreover, we shall also develop the
sizing and performance analyses that are necessary for the development of an overall system
architecture.
10.1 OPERATING SYSTEMS
The operating system is that software that allows all of the different processors to be
cholecystitis utilized to their best degree. In particular, the operating system will allow for
control of the memory of the system and the files that are to be managed by the processors. It
also allow for the overall management of the processes that are running on the system. Finally,
the operating system allows for the control of the input and output devices on the system. In a
distributed environment, the operating system can be extended to include all of the processes in
the system and in turn be the manager of the overall files in the system. As we discussed in the
323 | P a g e
development of the distributed database environment, the distributed operating environment,
builds upon the stand alone system.
In this section, we first develop an understanding of the standalone operating system and then we
develop the overall concepts of the distributed operating system. In particular we develop a
system called, MEDOS, standing for the Media Distributed Operating System. This system has
been developed for use in a fully distributed multimedia environment.
7.1.1 OS Structures
An operating system is simply a collection of software that, at one side controls the hardware and
other external assets of the computing environment, and at the other side, provides a ready means
for the user of the resources to incorporate them into the overall application. Thus we can view
the operating system from two directions, first that towards the resources managed, and second
towards the end users need to manage the overall set of resources. There are generally five major
functions of an operating system as we understand it today. These are:
1. Process Management
2. Memory Management
3. I/O Management
4. Device management
5. ,
These five major functions relate to typically a single standalone environment that focuses on the
need to manage the resources for a single machine. As we have discussed in the development of
multimedia communications, the primary element is that of the session. In contrast, in the
environment of the single machine, even with multiple users and multiple applications, the
primary focus is the process (an program in execution). There are significant differences in the
ability to meet the needs of a process as compared to a session. In this section, we shall review
the structure of current operating systems, and expand the concept to the area of a fully
distributed operating systems focused on multimedia applications.
Operating systems have evolved from the simplest ones that were basically the operator
themselves, through batch systems, and upwards to today's systems that have the capability to
perform multi-processing and multi-tasking. The development of such common operating system
environments such as UNIX and MS-DOS are responses to the end users' needs to manage the
324 | P a g e
computing resources of the specific location. Expanding the resource management to multiple
locations has been achieved through communications channels that we have discussed in the last
chapter.
To provide an oversight, we can view operating systems to fall in four general categories. These
are a local operating system (LOS), a networked operating system (LOS), a distributed operating
system (DOS) and a multi-media distributed operating system (MEDOS). Our objective in this
section is to develop the structure of a MEDOS and to do so by showing how it is a natural
evolution of the LOS environment. Before continuing, let us first define in some detail each of
these separate operating system environments, with emphasis on the requirements of the end user
that they satisfy.
Before comparing the different architectures for the operating system environments, let us first
consider a canonical environment of an operating system. The Figure below shows the canonical
structure for the operating system. It consists of six levels. These levels are:
Shell
Process Initiation Process Completion Create/Destroy Process Process Message Mgt
Kernel
Foreground
Background
Kernel
Unit Diagnostics Maintenance and Control
Shell
SOH I/O Mgt File Mgt Supervise Rpt Gen Monitor/Control
Another view of such a structure is shown below containing more enhanced elements is as
follows.
325 | P a g e
Hardware
Process Scheduling Memory Management Synchronization
Memeory Release Memory Allocation
I/O System Mgt I/O Initiation I/O Scheduler
Open/Close File Create/Destroy File Read/Write File
I/O Display
Command Manager OS Command Interface
The specifics as above are defined as follows:
1. Hardware: This is the innermost level and represents the totality of all hardware that the
operating system could manage as resources to the end user and the application. Typical
of such resources could be the processor or processors involved in the computation, the
memory units, printers, displays and entry devices for the end user.
2. Generics: This typically is microcode that is written specifically for each device and may
also typically be provided by device vendors for the local access to interface and manage
the individual devices. For the most part, the device drivers. And fault monitors are part
of the generic elements.
3. Kernel: This is the first part of the operating system. The kernel element, for our
purposes, is the layer of operating systems functions that directly interface with the
devices and hardware. In this canonical structure, all kernel functions of the operating
system are hardware directed, as compared to the shell functions that are applications
support directed. Thus memory management is a kernel layer function whereas file
management is a shell layer function.
4. Shell: The shell is the second layer of the operating system and is that layer that looks out
towards the user. It provides for the user interface and also allows for direct access by the
326 | P a g e
higher layers to all the service provided by the operating system. Access to the shell layer
is through a set of commands that we call primitives. Primitives have a defined syntax
and can be used by the end user to not only access all the operating system facilities but
to manage them in some detail.
5. Services: The services layer is provides for special support functionality to the end user
application. Typical service layer functionality is that of a windowing environment that
evokes the functionality of the operating system. X Windows is a typical service layer
functionality that supports the capabilities in UNIX base system. Other services could
include the ability to perform sessioning, the ability to mail in the network, the ability to
perform communications service between various nodes ijn the network and other such
services. The service layer can also be thought of as the OSI layer interface at the
applications layer.
The layered approach to an OS is shown below:
Applications
Services
Applications
Shell
Kernel
Generics
HW
Typical services that can be supplied by the system are as follows:
1. Communications: This service allows for the connecting of a local environment to any
other local environment. This service usually utilized the ISO based layered approach to
establish the communications link.
327 | P a g e
2. File: This service attends to the need of the end user in accessing and maintaining their
overall files. In a distributed environment the file service extends to sharing in a common
and seamless fashion all of the important file elements.
3. Directory: This service allows access to the location and names of all files on the system.
If we consider a file in the extended sense of UNIX then we can see that the directory
provides the overall access mechanism for any readable and/or operable entity or agent
on the system.
4. Mail: This service allows for the sending of data elements from one user to another on the
system.
5. Session: This is and advanced service that allows for the establishment and maintenance
of a session based service. We shall be discussing this at length.
6. Gateway: This service allows for the interconnection of multiple users to other networks.
It extends the communications service by providing a level of protocol translation and
conversion.
7. Presentation: This service consists of means and methods to present information on the
end users console. An example, as we have discussed is the X Windows service that is
used for bit mapped presentation.
8. Print: This service allows for the output in a hardcopy format of any of the data or
processed files in the system.
These services are only examples of the many that can be generated at the services layer. The
services are accessed via Service primitives that allow for specification of the details of the
service elements. In turn the services are generated by using a collection of the Shell Primitives.
Applications: This is the layer that include the set of programs that make up the end user
application. The applications may be one or many, and they may be associated with one or many
end users.
We can now use this canonical model and explain each of the four types of operating systems in
terms of the canonical form. The reader should note that we have extended the models for the
operating system environment beyond what may generally appear in the literature. The classic
text by Madnick and Donovan provide an excellent review of operating systems in the mid-
328 | P a g e
1970's. The more recent texts by Tannenbaum and by Deitel provide excellent updates to the
environment.
We consider now the different types of OS such operating systems:
10.1.1 Local Operating System (LOS)
A Local Operating System, LOS, is the result of the evolution of the end users need to readily
access many of the system resources and the systems designers needs to allow for maximum
flexibility in explaining both the applications run on the machine and the number of users
utilizing the machine resources. From the end user's perspective, there is a need to access files
that are stored in the system and to manipulate those files for the purpose of a specific
application. We allow the access to the LOS through a user interface that is a command
interpreter. The commands that allow the user to manipulate files are similar to commands that
permits all of the functions in the LOS. We shall call these commands the operating system
primitives, or primitives for simplicity.
The LOS focuses on the local environment and its resources. It allows for the input and output of
the end users commands and data and also manages all other system resource I/O. In a similar
fashion to file management, it manages the system memory, and if necessary can interact with
the management of processes to permit certain memory to be brought to faster processing
locations for better optimization of applications run time. Typical of this latter approach is a
software controlled cache function.
To the user of an MS-DOS system, the commands such as dir ( for directory), path, and file, are
typical of the file management commands readily available.
We depict the structure of an LOS operating an application that may share one or more
machines. The application is bound between the two machines by the services layer
communication service. This means that the machines are totally separate and are self-contained.
Only in an abstract sense are the applications shared. In reality they are fully resident in each
machine.
329 | P a g e
LOS
Applications
Services
Services
Shell
Shell
Kernel
Kernel
Generics
Generics
HW
HW
10.1.2 Network Operating System (NOS)
A Network Operating System is the next step up in complexity from the local OS. In the local
OS environment, all of the resources are local and are owned and operated by the single OS. In
the network OS case, there are multiple locations, but all of the resources are locally owned and
controlled. Thus if a user desires to use the resources at another location, that user must log onto
those resources through the network.
We can see below in the NOS case, the OS elements are still discrete even though the users may
be sharing the applications and now Rtes services layer. The NOS structure allows access to all
elements through a shared network communications protocol and also a set of other shares
service elements.
NOS
Applications
Services
Shell Shell
Kernel Kernel
Generi Generi
cs cs
HW HW
330 | P a g e
10.1.3 Distributed Operating System (DOS)
In a Distributed Operating System, only the hardware and its associated generics are still
associated on a logical basis with the individual sites. In this case the operating system functions
of both the shell and kernel are shared amounts all the users. In a fully implemented DOS
environment, no user will be able to identify the specific resource being utilized at any time. The
sharing of all of the resources and their access is through the common distributed shall/kernel
combination. Below shows how such a DOS environment can be envisioned. One could
conceive of the day when a fully distributed system is developed with sharing of even the
generics and hardware. Such is not the issue in the present discussion.
DOS
Applications
Services
Shell
Generics
Kernel
HW
Generics
Generics
HW
HW
10.1.4 Media Distributed Operating System (MEDOS):
We have discussed the concept of a distribute operating system and the focus has been on the
system that has standard data elements and storage. In a multimedia environment, the hardware
the lies in between the system is more complex and hardware elements are multimedia in
character. Thus the system must combine all of the elements that we have discussed previously.
The overall MEDOS architecture is shown below:
331 | P a g e
Shell
Session Mgt Session Establish Agent Monitor Shell/Kernel Interface
Kernel
Interprocess Control Transport Mgt Resource Mgt I/O Drivers
Generics
Foreground
HW
Background
Kernel
Network Diagnostics Maintenance and Control
Shell
SOH Billing Usage File and DB Mgt Supervise Rpt Gen Monitor/Control
The following is the Kernel Architecture:
Shell Bandwidth
Kernel Manager
Shell Kernel
IPC Transport I/O Driver
Interface
Foreground
Background
Network
SOH Maintenance
Diagnostics
The following is the Shell Architecture:
332 | P a g e
Session
Establishment
Session Shell Kernel

Agent Monitor
Manager Interface
Foreground
Background
Database Report
SOH Supervisor
Loader Generator
Maintenance
Billing and Control
Database
The MEDOS system was a multimedia directed distributed system that worked across a multiple
set of platforms and data sets using a high speed, 45 Mbps, channel when it was deployed in
19893.
10.2 OS DETAILS
Operating Systems have been developed to provide serves to a wide variety of machines and in a
large set of environments. Typically, there are four major functions of an operating system (see
Madnick and Donovan):
1. Memory Management
2. File management
3. Process management
4. I/O management
We have also included a fifth function (see Tannenbaum) :
3
Note that the author had architected and named the system in late 1986 and in three years of debelopment it was deployed for test and
evaluation.
333 | P a g e
5. Device management
Very simply, these five functions perform the following tasks.
Memory Management: This function manages the reel and non-real time memory of the system.
It allocates space in memory and moves date to and from different parts of the systems memory.
As we look towards multimedia applications, memory management becomes even more complex
than that for standard digital data. The nature of the storage media changes and the use of real
and virtual storage becomes more complex.
File management: A file is defined as a collection of data that has a specified name (see Deitel).
In contrast to memory management, which manages the physical resources of the storage
devices, file management manages the allocation of data to memory that is used in the context of
specific applications. As with memory management, file management for multimedia
applications is more complete because of the need to create files of often unbounded and
disparate media.
Process management: A process is a program in execution. The management of processes entails

the allocation of system resources to ensure the effective execution of all processes. The
implementation of interprocess control (IPC) for the coordination and swapping of processes in
execution is also a key ingredient of this function.
I/O Management: The management of the various input and output devices is a key element of
this function. This includes the management of printers, display devices and other elements.
Device management: Device management typically refers to internal elements of the

computation process. It may often be bundled into the management of I/O elements.
These five areas are the classical elements of the operating system structure. However, below,
we have displayed the canonical structure of an operating system. This canonical structure shows
two major divisions. The first division is that between the shell and the kernel. As we have stated
before, the shell is that part of the operating system that relates outwardly to the end user, and
specifically to the services layer. The kernel looks inward to the devices. The second division is
in terms of foreground and background process. The foreground represents real time or near real
time processes that operate in the OS. The background processes are no real-time processes.
The division of an operating system into these two divisions is an important one from an
operational perspective. The shell/kernel division allows the designer to focus on the directional
emphasis on interfacing and functionality. The foreground/background partition focuses on the
real time nature of the associated processes.
334 | P a g e
In the canonical structure, we have shown sixteen separate processes that are used to define the
canonical operating system. Theses represent those that are normally a part of the standard five
elements plus a set that we have added for the purpose of expanded beyond the local OS to a
fully distributed environment. We shall now discuss the contents of each of the processes.
1. Shell
a. Foreground
i. Process management
ii. Process establishment
iii. Agent Monitor
iv. Shell/Kernel Interface
b. Background
i. State of Health
ii. I/O management
iii. File Management
iv. Supervisor
v. Report Generation
vi. Maintenance and Control
2. Kernel
c. Foreground
i. Media (Memory) Management o
ii. Communications Interface o
iii. Resource Management o
iv. I/O Drivers o
d. Background
i. Unit Diagnostics
ii. Maintenance and Control
Figure Operating System Structure: Canonical Example
335 | P a g e
Applications End user applications software
Services MM session oriented support
Shell OS/System interface services
Kernel OS/Peripherals interface services
Generics HW/Peripherals support SW
HW HW elements
We can now use the discussion of the canonical operating system and show how existing
operating systems map into the canonical model. In the Figure, we have presented three of the
more common operating systems in use, VMS, UNIX and MS-DOS.
10.2.1 Specifics
Now that we have defined the overall canonical model for an operating system, and we have
provided a comparison to specific existing operating systems, we can now demonstrate how each
of the key processes is developed with module specification diagrams.
The operating system structure that we developed in the last section was a generic structure that
can apply to all possible types of operating systems. However, as we have seen in the examples
of existing operating systems that are local in nature, many of the elements of the canonical
model are not needed in the system. What distinguishes a distributed operating system from a
local system is the need to ensure that all of the physical resources are treated as one set of local
resources and not just as a collection of separate local sets. Thus, for a distributed operating
system, there is a need for significantly increased communications management amounts the
physical resources.
Let us begin by considering a simple example in the area of memory management. In a local
operating system, we can partition the memory and manage it from a single local point. Let us
assume that the system has both RAM memory and Disk memory and that we page memory
from disk into RAM for processing purposes. Let us assume that main memory (eg RAM) has
capacity of CH and that there are N processes that are being run simultaneously.
336 | P a g e
Let Pi represent the ith process. Let Pi take memory space MPi and let the data requested for Pi
take space MMi. Thus using an appropriate swapping algorithm, we can use a performance
measure of the system and the algorithm that measure the efficiency of the usage of main
memory and the efficiency of the time efficiency of use. We call this the machines time-
bandwidth product. It is defined as:
BT = (MEMeff * MIPS)*(Timeeff* Tcycle)
where:
MEMeff = the percent of main memory fillable on average by the swapping algorithm.
MIPS = the processing rate of the machine in instructions per second.
Time eff = the percent of a cycle time that is used for processing as compared to all other factors
such as data transfer etc.
T cycle = the average instruction cycle time
Now the T eff can be calculated as follows. If T load is the total time to load a main memory
cell, and T proc is the total processing time for the command of that process in that cell then we
define T eff as:
T eff =T proc
Now generally we have:
T load « T proc
because the memory access time is significantly short. The access time of the memory is
generally short and is dependent mainly upon the I/O characteristics of the memory device. In a
distributed environment however, the access time is increased due to the communications
associated with the memory call plus the nomad memory I/O access time. Specifically we have:
T load = T I/O + T comm
Now when we design the paging algorithm, or consider the need for virtual memory, even in one
machine, we must include the communications effects into the analysis. Frequently:
337 | P a g e
T comm » T I/O
Let us calculate some of the numbers in these equations to give the student a feeling for the
dimensions of the problem. The speed of light is about one foot per nanosecond. If we have a
distributed system with separation of 500 miles on average, then this is 2.5 10 6 feet and equals a
delay of 2.5 msec. If the I/O to the memory unit works at microsecond speed, then we have the
inequality indicated with three orders of magnitude difference. In addition if we have a 2 MIP
machine, and we page in 100 instructions per second, the processing time is 50 microseconds.
Clearly T proc is greater than T I/O but much less than T comm. It will be this factor that we
shall see dominating the DOS performance.
Table 7.x depicts the mapping of the detailed operating systems functions as described against
the canonical model and shows how they relate to the higher level functions commonly used in
operating system descriptions. We shall now use the more common structure to discuss the key
features that must be considered in a distributed operating system as compared to a local
operating system. Each of these issue relate to the problems that we have discussed in the above
example. That is, it is now possible to use other memory devices and other processors that are
connected more loosely and with greater delays than generally anticipated.
Let us now consider all of the major elements of the DOS and detail their key performance
factors and compare these against a standard LOS.
10.2.2 Process Management
The process management function is directed towards the overall management of the processes
that are operational in the system. It looks outward towards the processes that are part of the
overall applications program and ensures that they are scheduled and interlaced in an effective
manner. Process management takes the form of managing the multiple users in the system and
matches the users and their needs with the overall set of system resources. A process is taken in
and out of the main memory and assigned system resources depending upon the needs of the
process, the priority of the end user and the availability of the resources.
10.2.3 Local System Performance
In a local operating system environment, the issues of process management relate to the
following items.
1. Process Establishment:
2. Interprocess Communications:
3. Process Monitoring:
4. Process Messaging:
338 | P a g e
5. Process Scheduling:
6. Process Termination:
10.2.4 Distributed Systems Design Factors
In a distributed environment, we include all of the above items plus several that are in addition to
these. Moreover, we need to expand the functionality of the separate elements of the process
management function to ensure that processes are properly managed across several resources.
Process management in a distributed environment may take on several different aspects. In the
local environment, we have a set of processes that can be existing on a single processor. They are
managed by the operating system of the local processor. In a distributed system, we can envision
several scenarios:
1. Single process per Processor/Multiple Processors: In this case there is a single process
that is operational at a single processor and it must communicate with other processes at
other processors.
2. Multiple Processes per Processor/Multiple processors: In this case we have multiple

processes that are operational at each processor but that they all must communicate
between themselves.
3. Single process per Multiple processor/MP: In this case we have a single process that is
shared amounts several processors. These processors are separated by many local
distances and thus have the dynamics associated with separate locations.
4. Multiple processes per Multiple Processors (MP): In this case we have multiple
processes, all communicating amongst themselves, distributed in multiple processors.
Process management in a distributed environment is increased in complexity by the sharing of

resources that are at disparate locations but the advantage is the ability to share resources and to
dynamically reallocate resources amongst many users.
10.2.5 Device Management
Device management is also known as processor management. It is the physical complement of

the process management that we have just discussed in the above are a process management. The
processor management function basically relates to the assignment of processes to specific
systems resources or processors. The process management function relates to the intra-process
management functions, namely how does on manage the process itself and in relation to their
processes. The processor management function relates to how we manage the processes in
339 | P a g e
relation to their outside environment. The processor management function relates to the overall
need of the system to perform the following types of task:
1. Job scheduling of the processes to the processors.
2. Traffic control of tasks between and amongst the
3. Synchronization of the processors amongst themselves.
Device management for local systems and distributed systems is very similar. The allocation of
resources can be managed on a centralized or a distributed fashion. In either of these cases the
specific algorithms may be static or dynamic, and in turn may be time driven or event drive.
Thus, the type of processor manager may be characterized by the tuple:
{Control, Timing, Response}
where the control is central or distributed, the timing is static or dynamic, and the response is
event or time driven. Thus there are the following types of processor management systems:
{Static, Dynamic, Event} etc.
This yields eight possible configurations of process management. We develop several algorithms
for the management under these domains in both the examples in this section and the problems.
In a distributed environment, the same set of variables apply, and moreover the same set of
algorithmic devices or artifacts are applicable.
10.2.6 I/O Management
The I/O management entails both the I/O to devices and peripherals as well as I/O to the end
user. We shall focus on both of this element. The physical device I/O is controlled by a set of
device drivers that talk directly to the device and are controlled by the management portion of
the I/O element. The end user I/O are interfaced by what we term primitives, which are
command sets that use a defined syntax and protocol and allow the user to interact with the
operating system directly.
10.2.7 Memory Management
Memory management provides the management of the physical memory of the system. There are
typically many forms of memory that are used in a commuting system. These range from the
local RAM memory storage, to local disk and local tape memory. Against these physical storage
340 | P a g e
media is the allocation of the memory that the process recognizes as important and the memory
as allocated and managed by the operating system. The process frequently has need to access
many elements of the physical memory and it then needs to changes and redirect that memory as
it is processed. The operating system must assist the process in its process management and
memory management functions in allocating this memory to be effectively used by the system.
Typically we have a system with a fixed amount of memory that is immediately accessible by the
process in operation. Let us defined each process by Pi and let the total memory be given by MT.
Let the set P{I,k} represent the process set at time k, and let this equal:
P{I,k} = { Pi : i I and kT< t < (k+l)T }
10.3 DISTRIBUTED PROCESSORS
In any distributed environment there are not only issues of the distributed data base and the
distributed operating system, but there is also the issue of the distributed processors and their
interconnection. This section addresses some of the issues of distributed processors.
Processor Types
Processor Performance
Processor Interconnection
10.4 DISTRIBUTED PROCESSES
As we recall from our earlier discussions a process is a program in execution. A program calls
upon all of the resources of the operating system, the databases and the processors. Thus there is
a significant different in a multimedia distributed environment for the functioning of a set of
distributed processes.
10.5 CONCLUSIONS
This chapter developed the overall distributed environment and discussed the key elements of
that environment. What should be clear after developing these many issues is that there still are
many open questions in distributed environments and these questions are driven by the
availability of new technology that allows for interconnection at extremely high data rates. In the
past, as we have discussed in the presentation of the ISO model, we have found that the data
rates were slow and that it was necessary to provide a significant overlay of protocols to assure
end to end performance. A distributed environment in that world was limited to slower acting
341 | P a g e
systems. In the world of unlimited bandwidth as presented by fiber links, a fully distributed
architecture can be developed.
342 | P a g e
11 CONCLUSIONS
In this book, we have developed a view of multimedia communications that combines the
structure of the underlying images and the storage and transport media along with a careful
understanding of the end user and their interactions with the different forms of media. Unlike the
world of computer communications where the computer can be trained to perform certain tasks
that are required for proper performance, the human in multimedia application is operating in a
mode of total creativity. This creativity results in complex and often unpredictable interactions
with the other human users. This interaction is the essence of the multimedia communications
theory.
We have tried in this book to present a canonical structure to view the multimedia multiuser
environment, focusing on the concept of the session. We have not focused on the psychological
inferences and actions of the end user in great detail. This is the artistic side of multimedia
communications and expands beyond just knowledge engineering or even epistemology.
11.1 KEY ISSUES
This book has developed a philosophy of multimedia communications that tries to separate the
"artistic" element from the engineering element. It may be criticized that such a separation is
difficult if not impossible, but our intentions were to gain a better focus on the system ideas by
quantifying the basis elements and not serializing them. Thus the approach has been alone the
more common engineering lines, understanding that the system definition and design are the key
characteristics. We can now review the overall key issues that we have developed in this book
and present what has been developed in context.
Multimedia communications, as we have seen, deals with the handling, manipulation, processing
and transmission of complex images that are highly interactive in nature. The insensitivity factor
focuses on the "conversational" mode of communications that we have discussed.
1. The Overall System Approach

2. Characterization
3. Performance and Sizing
4. Feel and Form versus Function and Formalism
5. New Paradigms of Presentation
6. Displaced Working Environments
7. End User as Design Element
8. Sessions
343 | P a g e
9. Distributed Environments
The evolution of multimedia communications will be enhanced by the development and

deployment of broadband communications as well as the development of smarter and more
sophisticated end user terminals. The most difficult problem is the issue of how do we effectively
interface with the human end user. Specifically what common paradigms are most effective in
what environments.
The trends that impact this area of multimedia communications are quite simple to articulate at
this stage but they may change as we enter into a time of rapid technological change.
Specifically, if we look at the model of the multimedia environment, and we map prognosticated
changes in each of these areas against the separate elements of the multimedia domain, we can
carefully determine the impact of the multimedia environment and determine the returns to that
environment.
Let us focus on the specific multimedia areas and discuss the trends in each of those areas that
will impact the development of an improved and more advanced multimedia communication
system.
These trends provide for a brief introduction of some of the options that are available to the user
and developer of multimedia communications networks. They present a view of what
is possible but do not present what will actually occur. We have developed some of these issues
based upon what is difficult to do in today's environment but new technology will most certainly
alter this view.
The discussion in the past section turns on views of future trends that are natural follow on to
what we have discussed in the body of the text. In this final section of the book we wish to share
some of the future research directions that are open and whose solutions provide the most fruitful
ground for future work.
1. Multimedia Database Structures
2. End User Interface Language Design
3. WAN Communications Architectures
4. Distributed Operating System Optimization
5. Overall System Optimization Performance
344 | P a g e
6. Multi Sense Interfaces
11.2 SUMMARY
This book presented a vision of multimedia communications as more than just an amalgam of
separate parts, disembodied from the ultimate user. It presents a holistic view that recognizes the
impact of the user on the services and the underlying technology. Multimedia communications is
a concept that is in its infancy but that also offers the user a significant increase in performance,
functionality and the ability to disembody themselves from the medium of communications.
As was stated earlier in the book, Marshall McLuhan stated that the introduction of a new
medium changed not only how we presented information, but ultimately what was considered
knowledge. This is both a challenge and a warning. The technological concepts presented in this
book provide for a significantly different view of media or presentation technology. It raises the
question of how best to interface the end user to a network that responds in near instantaneous
speed at almost all data rates. It, in effect, introduces a plethora of new media, thus challenging
and warning us simultaneously.
345 | P a g e
12 REFERENCES
1. Abbatiello and Sarch, A Typology of Local Area Networks - Confused by the array of
LANs, some users fall back on switching, in Telecom m unications and Data Communications
Factbook Data C om m unications and C C M I/McGraw -H ill, 1987
2. Aguietal, A Painting Algorithm of Monochromatic Images Using Color-Sequences,
translated from Denshi Tsushin Gakkai Ronbushi, vol 67-D, No.3, March 1984, Scripta
Publishing Company, 1984
3. Alemany, R. Kasturi , A Computer Vision System for Interpretation of Paper-Based
Maps in Applications of Digital Image Processing X - Proceedings, edited by A.G. Tescher SPIE
- The International Society for Optical Engineering, 1987
4. Arbib, M.A., A. R. Hanson, Vision, Brain and Cooperative Computation, MIT Press
(Cambridge, MA), 1990.
5. Arms, C., Campus Networking Strategies, DIGITAL Press (Maynard, MA), 1988.
6. Atkins, P. Physical Chemistry, Freeman (New York) 1990.
7. Avery, J.E., VAN, The Regulated Regime, Telecom and Posts Div, Dept of Trade and
Industry, UK, Conf on Communications, IEE, pp 78-81, May 1986.
8. Ayer, A.J., Philosophy in the Twentieth Century, Vantage (New York), 1984.
9. Ayer, A.J., Russell, The Wodburn Press (London), 1974.
10. Ayer, A.J., Wittgenstein, Random House (New York), 1985.
11. Barlow, W. The Broadband Revolution, Info Tech and Pub Policy, Vl 8, No 1, pp 6-8,
1989.
12. Barlow, W. The Broadband Revolution, Info Tech and Pub Policy, Vl 8, No 1, pp 6-8,
1989.
13. Barrett, W., The Illusion of Technique, Anchor Press (New York), 1978.
14. Bartel, B., S. Matsuda, Seeing Red, Science, Vol 299, 17 Jan 2003, pp 352-353.
15. Bell Labs, Transmission Systems for Communications, fifth edition Bell Telephone
Laboratories Incorporated, 1982
16. Bell, D., The Coming of the Industrial Society, Basic Books (New York), 1973.
17. Bell, T., Technical Challenges to a Decentralized Phone System, IEEE Spectrum, pp 32-
37, Sept.,
18. Bell, T., Technical Challenges to a Decentralized Phone System, IEEE Spectrum, pp 32-
37, Sept., 1990.
19. Bender et al, CRT Typeface Design and Evaluation, MIT Industrial Liaison Report
S0288-069,1988
20. Bender, Anti-Aliasing for Broadcast Compatible Television, SID 84 Digest
346 | P a g e
21. Bender. Anti-Aliasing for Broadcast Compatible Television, SID 84 Digest, 1984CRT
Typeface Design and Evaluation, W. Bender, R.A. Crespo, PJ. Kennedy, R. Oakley MIT
Industrial Liaison Report S0288-069,1988
22. Berns, R. S., Principles of Color Technology, Wiley (New York) 2000.
23. Bernstein, J., Three Degrees Above Zero, Scribners ( New York), 1984.
24. Blackwood, M.A., A. Girschick, Theory of Games and Statistical Decisions, Wiley (New
York),
25. Blackwood, M.A., A. Girschick, Theory of Games and Statistical Decisions, Wiley (New
York), 1954.
26. Born, M., E. Wolf, Principles of Optics, 4th Ed, Pergamon (New York) 1970.
27. Bostwick, W.E., Program Plan for the National Research and Education Network, Dept.
Of Energy,
28. Bostwick, W.E., Program Plan for the National Research and Education Network, Dept.
Of Energy, May, 1989.
29. Bradley, Alan, Optical Storage for Computers, Wiley (New York) 1989.
30. Brealey, R., S. Myers, Principles of Corporate Finance, McGraw Hill (New York), 1990.
31. Burns, Alan, Andy, Wellings, Real Time Systems and Their Programming languages,
Addison Wesley (Reading, MA), 1989.
32. Business Review, pp. 114-120, September-October, 1990.
33. Bux, Token-Ring Local Area N etw orks and Their Performance Proceedings o f the
IEEE, vol.77, no. 2, February 1989
34. Byrne et al, Broadband ISDN T echnology and Architecture IEEE Network, January,
1989
35. Carnevale, M.L., "Untangling the Debate over Cable Television", Wall Street Journal, p.
B1, March, 19, 1990.
36. Casavant, Analysis of Three Dynam ic Distributed Load-Balancing Strategies with
Varying Global Information Requirements 1987 IEEE 7th International Conference on
Distributed Computing System s Computer Society Press of the IEEE, 1987
37. Chang, N.S., K.S. Fu, Picture Query Languages for Pictorial Data Base Systems, IEEE
Computer, Nov 1981.
38. Chang, S., T. Kunii, Pictorial Data Base Systems, IEEE Computer, Nov 1981, pp 13-19.
39. Chiariglione et al, An Integrated Approach to Video Standard Conversion and Pre-
Processing for Videoconference Codec in Image Coding edited by M. Kunt, T.S. Huang
Proceedings of the SPIE The International Society for Optical Engineering, volume 594,1985
40. Chomsky, N., Aspects of the Theory of Syntax, MIT Press (Cambridge, MA), 1965.
41. Christman et al, Which Kinds of OS Mechanism s Should be Provided for Database
Management? Experiences w ith Distributed Systems edited by G. G oos and J. Hartmannis
Lecture Notes in Computer Science, Springer-Verlag, 1987
347 | P a g e
42. Christodoulakis, et al, The Multimedia Object Presentation Manager of MINOS, ACM,
1986, pp 295-310.
43. Christodoulakis, S., et al, Design and Performance Considerations for an Optical Disk
Based Multimedia Object Server, IEEE Computer, Dec 1989, pp 45-56.
44. Chung and Pereira, Timed Petri Net Representation of SMIL, IEEE Multimedia, 2005
45. Churchland, P.S., Neurophilosophy, MIT Press (Cambridge, MA), 1986.
46. Ciardo et al, A Characterization of the Stochastic Process Underlying a Stochastic Petri
Net, IEEE Transactions On Software Engineering, Vol. 20, No. 7, July 1994
47. Clark, A., Microcognition, MIT Press (Cambridge, MA), 1989.
48. Coll, S. The Deal of the Century, Atheneum (New York), 1986.
49. Copeland, T.E., J.F. Weston, Financial Theory and Corporate Policy, Addison Wesley
(Reading, MA) 1983.
50. Cortadella et al, Synthesizing Petri Nets from State-BasedModels, IEEE, 1995
51. Couch, S., T.P. McGarty, H. Kahan, QUBE: The Medium of Interactive Response, A
Compendium for Direct Marketeers, Direct Mktg Assn., pp 162-165, 1982.
52. de Sola Pool, I., Technologies Without Barriers, Harvard University Press (Cambridge,
MA), 1990.
53. de Sola Pool, I., The Social Impact of the Telephone, MIT Press (Cambridge, MA), 1977.
54. Delbruck, Max, Mind From Matter, Blackwell (Palo Alto, CA), 1986.
55. Depew, D.J., B.H. Weber, Evolution at a Crossroads, MIT Press (Cambridge, MA), 1985.
56. Dertouzos, M.L., et al, Made in America, MIT Press (Cambridge, MA), 1989.
57. Dertouzos, M.L., J. Moses, The Computer Age, MIT Press (Cambridge, MA), 1979.
58. Dorfman, R.A., P.A. Samuelson, R.M. Solow, Linear Programming and Economic
Analysis, Dover (New York), 1986.
59. Dreizen, Content-Driven Progressive Transmission of Grey-Scale Images IEEE
transactions on Communications, vol. COM-35, no.3, March 1987.
60. Dreyfus, Hubert L., Being-in-the-World, The MIT Press (Cambridge, Mass.), 1991.
61. Drucker, Peter F., Adventures of a Bystander, Harper Row (New York), 1979.
62. Dugan, D.J., R. Standard, Barriers to Marginal Cost Pricing in Regulated
Telecommunications, Public Utilities Fortn., vol 116, No 11, pp 43-50, Nov 1985.
63. Eco, U., A Theory of Semiotics, Indiana University Press (Bloomfield, IN), 1979.
64. Egan, B.L., Costing and Pricing of the Network of the Future, Proc of International
Switching Symposium, pp 483-490, 1987.
65. Egan, B.L., T.C. Halpin, The Welfare Economics of Alternative Access Carriage Rate
Structures in the United States, Telecom Journal, Vol 54, No 1, pp 46-56, Jan 1987.
348 | P a g e
66. Elmarsi, Ramez, Shamkant Navathe, Fundamentals of Database Systems, Benjamam (
Redwood City, CA) 1989.
67. Fleming, S., What Users Can Expect From New Virtual Wideband Services,
Telecommunications, pp 29-44, October, 1990.
68. Franklin, Problems with Raster Graphics Algorithms in Data Structures for Raster
Graphics edited by L.R.A. Kessener, F J. Peters and M.L.P. van Lierop Eurographic Seminars,
Springer-Verlag, 1986
69. Freeman, The 35mm Handbook, Ziff David Books, 1980
70. Fruhan, W.E., Financial Strategy, Irwin (Homewood, IL), 1979.
71. Fulhaber, G.R., Pricing Internet: Efficient Subsidy, Information Infrastructures for the
1990s, J.F.
72. Gadamer, Hans Georg, Philosophical Hermeneutics, U. Cal Press (Berkeley), 1976.
73. Gadamer, Hans Georg, Truth and Method, Crossroad (New York), 1990.
74. Gadamer, Hans-Georg, Philosophical Apprenticeships, MIT Press (Cambridge) 1985.
75. Gadamer, Hans-Georg, Reason in the Age of Science, MIT Press (Cambridge), 1981.
76. Gawdun, M., Private-Public Network Internetworking, Telecommunications, Vol 21, No
11, pp 49- 58, Nov 1987.
77. Gechter and OReilly, Conceptual Issues for ATM IEEE N etw ork, January 1989
78. Geller, H., US Domestic Telecommunications Policy in the Next Five Years, IEEE
Comm Mag, Vol
79. Geller, H., US Domestic Telecommunications Policy in the Next Five Years, IEEE
Comm Mag, Vol 27, No 8, pp 19-23, Aug 1989.
80. Ghafoor et al, An Interconnection T opology for Fault-Tolerant M ultiprocessor System s
1987 IEEE 7th International Conference on Distributed C om puting
81. Goos and Hartmanis, Communication Support for D istributed Processing : D esign and
Im plem entation Issues L. Svobodova in Networking in O pen System s Lecture Notes in
Computer Science, Springer-Verlag, 1987
82. Grayling, A. C., Wittgenstein, Oxford (Oxford) 1988.
83. Green, Implementation of FDDI : a 100 M bit Token Ring in Localnet 86 Online
Publishers, 1986
84. Hall, The Application of Human Visual System Models to Digital Color Image
Compression in Image Coding Proceedings of the SPIE - The International Society for Optical
Engineering, volume 594,1985
85. Haller, Rudolf, Questions on Wittgenstein, University of Nebraska Press (Great Britian),
1988.
86. Handel, Evolution of ISDN tow ards Broadband ISDN IEEE Netw ork, January 1989
349 | P a g e
87. Hanson, Owen, Design of Computer Data Files, Computer Science Press (Rockville,
MD) 1988.
88. Hasegawa ey al, Com m unication-System Models Embedded in the OSI-Reference
Model, a Survey IFIP, North-Holland, 1986
89. Heidegger, M., Basic Writings, Harper & Row (New York), 1977.
90. Heidegger, Martin, An Introduction to Metaphysics, Yale (New Haven) 1959.
91. Heidegger, Martin, Being and Time, Harper & Row (New York) 1962.
92. Heidegger, Martin, Early Greek Thinking, Harper & Row (New York) 1979.
93. Heidegger, Martin, On Time and Being, Harper & Row (New York) 1972.
94. Henderson, J.M., R.E. Quandt, Microeconomic Theory, McGraw Hill (New York), 1980.
95. Hills, J., Issues in Telecommunications Policy- A Review, Oxford Surveys in
Information Technology, Vol 4, pp 57-96, 1987.
96. Hoshiko et al, A l00 Mb/s Optical Token Ring N etw ork Suitable for H igh-Speed Inter-
Processor C om m unications 1987 IEEE 7*h International Conference on D istributed C om
puting System s edited by R. Popescu-Zeletin, G. Le Lann, K.H. Kim Com puter Society Press of
the IEEE, 1987
97. Hsing and Liou, The Challenge o f VLSI Technology to Video Communications in
Applications of Digital Image Processing X - Proceedings SPIE - The International Society for
Optical Engineering, 1987
98. Huber, P.W., The Geodesic Network, U.S. Department of Justice, Washington, DC,
January, 1987.
99. Hudson, H.E., Proliferation and Convergence of Electronic Media, First World Electronic
Media Symposium, pp 335-339, 1989.
100. Illich, I., B. Sanders, ABC, The Alphabetization of the Popular Mind, Vintage (New
York), 1988.
101. Irwin, M.R., M.J. Merenda, Corporate Networks, Privatization ans State Sovereignty,
Telecommunications policy, Vol 13, No 4, pp 329-335, Dec 1989.
102. Jackendoff, Ray, Semantics and Cognition, MIT Press (Cambridge) 1988.
103. Jansen. "Data Structures for Ray Tracing Data Structures for Raster Graphics edited by
L.R.A. Kessener, F J. Peters and M.L.P. van Lierop Eurographic Seminars, Springer-Verlag,
1986
104. Jantzen, H., High Gothic, Princeton University Press (Princeton, NJ), 1984.
105. Jayant and P. Noll, Digital Coding of Waveforms -, Principles and Applications to
Speech and Video, Prentice-Hall Signal Processing Series, 1984
106. Jenkins, F. A., H. E. White, Fundamentals of Optics, McGraw Hill (New York) 1957.
107. Johnson, Design Issuses in Netw ork in g Incompatible O perating System s Localnet 86
Online Publishers, 1986
350 | P a g e
108. Judd, D. B. et al, Color 2ND Edition, Wiley (New York) 1963.
109. Kaelin, E. F., Heideggers Being and Time, Florida State (Tallahassee) 1988.
110. Kahin, B., The NREN as a Quasi-Public Network: Access, Use, and Pricing, J.F.
Kennedy School of Government, Harvard University, 90-01, Feb., 1990.
111. Kahn, A.E., The Economics of Regulation, MIT Press (Cambridge, MA), 1989.
112. Kaiser and Morss, Direct Comparison of 35mm Film and High Definition Television, in
High Definition Television Colloquium 85, vol. 2, Minister of Supplies and Services, Canada,
1986
113. Kernel and Cherion, Request-R esponse am d M ulticast Interprocess C om m unication
in the in Networking in Open Systems Lecture N otes in Com puter Science, Springer-Verlag,
1987
114. Kim, W., H. Chou, Versions of Schema for Object Oriented Databases, Conf VLDB,
1988.
115. Kirstein, P.T., An Introduction to International Research Networks, IEEE Int Council for
Comptr Comm, Ninth International Conf, pp 416-418, 1988.
116. Knightsron The Reference M odel in Standards for O pen System s Interconnections
McGraw-Hill, 1988
117. Knoll and Depp, Adapative Gray Scale Mapping to Reduce Registration Noise in
Difference Images, in Computer Vision, Graphics and Image Processing 33, 129-137,1986.
118. Kohno, H., H. Mitomo, Optimal Pricing f Telecommunications Service in Advanced
Information Oriented Society, Proceedings of International Conf on Info Tech, pp 195-213,
1988.
119. Kolnik and Garodnick, First FDDI Local Area in Proceedings of the 1 2 ^ Conference on
Local Computer Networks of the Computer Society of the IEEE, 1987
120. Konsynski, B.R., E.W. McFarlan, Information Partnerships - Shared Data, Shared Scale,
Harvard Business Review, pp. 114-120, September-October, 1990.
121. Kraus, C.R., A.W. Duerig, The Rape of Ma Bell, Lyle Stuart (Secaucus, NJ) 1988.
122. Krishnamurthy, E.V., Parallel Processing, Addison Wesley (Reading, MA), 1989.
123. Kuhn, T.S., The Structure of Scientific Revolutions, Univ Chicago Press (Chicago),
1970.
124. Kung, Hans, Theology for the Third Millenium, Doubleday (New York), 1988.
125. Kunii, T. L., Visual Database Systems, North Holland (Amsterdam), 1989.
126. Kurzweil, R., The Age of Intelligent machines, MIT Press (Cambridge, MA), 1990.
127. Larmouth, Com m itm ent, Concurrency and Recovery - ISO-CASE and IBM LU.62 in
Networking in Open Systems edited by G. Goos and J. Hartmanis Lecture Notes in Computer
Science, Springer-Verlag, 1987
128. Lawrence, P.R., D. Dyer, Renewing American Industry, Free Press (New York), 1983.
351 | P a g e
129. Leger et al, Distributed Arithmetic Implementation of the D.C.T. for Real Time
Photvideotex on ISDN, Advances in Image Processing - Proceedings edited by A. Oosterlinck,
A.G. Tescher SPIE - The International Society for Optical Engineering, 1987
130. Levi, L., Applied Optics, Wiley (New York) 1968.
131. Lierop, Intermediate Data Structures for Display Algorithms in Data Structures for Raster
Graphics edited by L .R A . Kessener, F.J. Peters and M.L.P. van Lierop, Eurographic Seminars,
132. Liff, Color and Black & White Television Theory and Servicing, Prentice-Hall, Inc.,
Englewood Cliffs, 1985
133. Linge, David E., Philosophical Hermeneutics, University of California Press (England),
1977.
134. Little, T.D.C., A. Ghafoor, Synchronization and Storage Models for Multimedia Objects,
IEEE Journal on Sel Areas in Comm, April, 1990, pp. 413-427.
135. Loomis, Mary, Data Management and File Structures, Prentice Hall (Englewood Cliffs)
1898.
136. Luce, R.D., H. Raiffa, Games and Decisions, Dover (New York), 1985.
137. Lynch, M., S. Woolgar, Representation in Scientific Practice, MIT Press (Cambridge,
MA), 1990.
138. Lyons, J., Noam Chomsky, Penguin (New York), 1978.
139. Mac Cormac, Earl R., A Cognitive Theory of Metaphor, MIT Press (Cambridge) 1985.
140. Machizawa, M. Tanak, Image data Compression Based on Interpolative DPCM by Area
Decomposition, in, tranalated from Denshi Tsushin Gakkai Ronbushi, vol 69-D, no.3, March
1986, in Sato et al, Systems and Computers in Japan, vol. 17, no.12, 1986
141. Maier, David, The Theory of Relational Databases, Computer Science Press (Rockville,
MD) 1983.
142. Management, HIMSS Conference, San Francisco, Feb. 1991.
143. Mandelbaum, R., P.A. Mandelbaum, The Strategic Future of Mid Level Networks, J.F.
Kennedy School of Government, Harvard University, Working Paper, October, 1990.
144. Markoff, J., Creating a Giant Computer Highway, New York Times, Sept. 2, 1990.
145. Mayne, Netw ork Services, W ide Area N etw orks in Linked Local Area N etw orks,
second edition John W iley & Sons, 1986
146. McCarthy, T., The Critical Theory of Jurgen Habermas, MIT Press (Cambridge, MA),
1978.
147. McCulloch, W., Embodiments of Mind, MIT Press (Cambridge, MA), 1988.
148. McGarty, T.P., R. Veith, Hybrid Cable and Telephone Networks, IEEE CompCon, 1983.
149. McGuiness, Brian, Wittgenstein: A Life, The University of California Press (Great
Britian), 1988.
352 | P a g e
150. McLuhan, M., The Gutenberg Galaxy, Univ Toronto Press (Toronto), 1962.
151. McLuhan, M., Understanding Media, NAL (New York), 1964.
152. Mee, C. Dennis, Eric D. Daniel, Magnetic Recording, McGraw Hill (New York) 1988.
153. Menzes, An Interconnection N etw ork Supporting Ralational Join Operations 1987 IEEE
7th International Conference on D istributed C om puting System s edited by R. Popescu-Zeletin,
G. Le Lann, K.H. Kim Computer Society Press of the IEEE, 1987
154. Milgrom, L. R., The Colours of Life, Oxford (New York) 1997.
155. Mill, John S., et al, Utilitarianism and Other Essays, Penguin Classics (England), 1987.
156. Minzer and Spears, New Directions in Signaling for Broadband ISDN S. E. M inzer, D.R.
Spears in IEEE Com m unications M agazine, February 1989
157. Monk, Ray, Ludwig Wittgenstein: The Duty of Genius, The Free Press (New York),
1990.
158. Morita and Kataoka, Color Display System for High-Definition Television High
Definition Television Colloquium 85, vol. 2, Minister of Supplies and Services, Canada, 1986
159. Morrill, Jane, Multimedia, BYTE, February, 1990, pp. 200- 237.
160. Mounce, H. O., Wittgenstein, U Chicago Press (Chicago) 1981.
161. Muroyama, J.H., H.G. Stever, Globalization of Technology, National Academy Press
(Washington, DC), 1988.
162. Nadel, L. et al, Neural Connections and Mental Computation, MIT Press (Cambridge,
MA), 1989.
163. Nickerson, R.S., Using Computers, MIT Press (Cambridge, MA), 1986.
164. Nicolau, Cosmos, An Architecture for Real Time Multimedia Communications Systems,
IEEE Journal on Sel Areas in Comm, April, 1990, pp. 391-400.
165. Nikolau et al, Processor Performance. Allocation and Relocation of Processes in
Distributed Computer System s in Current A dvances in D istributed C om puting and Com m
unications Computer Science Press, 1987
166. Noam, E. M., Network Tipping and the Tragedy of the Common Network, J.F. Kennedy
School of Government, Harvard University, Working Paper, October, 1990.
167. OHara, S., The Evolution of A Modern Telecommunications Network to the Year 2000
and Beyond, IEE Proc, Vol 132, No 7, pp 467-480, 1985.
168. Osherman, Daniel N., Edward E. Smith, Thinking, MIT Press (Cambridge), 1990.
169. Osherman, Daniel N., Howard Lasnik, Language, MIT Press (Cambridge) 1990.
170. Osherson, Daniel N. et al, Visual Cognition and Action, MIT Press (Cambridge) 1990.
171. Parsaye, Kamran, et al, Intelligent Databases, Wiley (New York) 1989.
172. Pears, David, Wittgenstein, Harvard Press (Cambridge),1969.
353 | P a g e
173. Peterson, Petri Net Theory and the Modeling of Systems, Prentice-Hall, Englewood
Cliffs, 1981.
174. Pierkainen and Harwood, Segmentation of Color Images Using Edge-Preserving Filters,
in Advances in Image Processing and Pattern Recognition, edited by V. Cappellini, R. Marconi
North-Holland, 1986
175. Pindyck, R.S., D.L. Rubinfield, Microeconomics, McGraw Hill (New York), 1989.
176. Pizano, A. et al, Specification of Spatial Integrity Constraints in Pictorial Databases,
IEEE Computer, Dec 1989, pp 59-71.
177. Plaehn, PHIGS : Programmers Hierarchical Interactive Graphics Standards - A Giant
Step toward a Universal Graphics Standard, Byte, November 1987
178. Plompen et al, An Image Knowledge Based Video Codec for Low Bitrates, Advances in
Image Processing - Proceedings, SPIE - The International Society for Optical Engineering, 1987
179. Poopescu et al, , System s Com puter Society Press of the IEEE, 1987
180. Porter, M., Competitive Advantage, Free Press (New York), 1985.
181. Porter, M., Competitive Strategy, Free Press (New York), 1980.
182. Porter, M., The Competitive Advantage of Nations, Free Press (New York), 1990.
183. Public Utilities Fortn., vol 116, No 11, pp 43-50, Nov 1985.
184. Raisbeck, G., Information Theory, MIT Press (Cambridge, MA), 1963.
185. Rhodes et al, "Locally Optimal Run-Length Compression Applied to CT Images, IEEE
Transactions on Medical Imaging, vol. MI-4, no. 2, June 1985
186. Rhodes et al, "Locally Optimal Run-Length Compression Applied to CRT Images in
IEEE Transactions on Medical Imaging, vol. MI-4, no.2, June 1985.
187. Rorty, Richard, Philosphy and the Mirror of Nature, Princeton University Press,
(Princeton, NJ),
188. Rosenfeld and Kak, Digital Picture Processing, Second Edition, vol. 1, Computer Science
and Applied Mathematics, Academic Press, 1982
189. Rossi, B., Optics, Addison Wesley (Reading, MA) 1957.
190. Rousseau, Jean-Jacques, The Social Contract, Penguin Classics (England), 1968.
191. Russell, B., The Problems of Philosophy, Oxford University Press (Oxford), 1959.
192. Rutkowski, A.M., Computer IV: Regulating the Public Information Fabric, Proc of the
Regional Conf of the International Council for Computer Communications, pp 131-135, 1987.
193. Sato et al, Data Compression of Gray-Scale Images by Level Plane Coding tranlsted from
Denshi Tsushin Gakkai Ronbushi, vol 69-B, no.8, August 1986, Electronics and
Communications in Japan, Part 1, Vol. 71, n o.l, 1988
194. Schreiber, A Generalized Image Processing System in "Fundamentals of Electronic
Imaging SystemsSome Aspects of Image Processing Springer Series in Information Sciences,
354 | P a g e
195. Schreiber, A Simple Television System in Fundamentals of Electronic Imaging Systems
Some Aspects of Image Processing, Springer Series in Information Sciences, Springer-Verlag,
1986
196. Schrube and Yucesan, Transforming Petri Nets Into Event Graph Models, Proceedings
Of the 1994 Winter Simulation Conference
197. Sears, F. W., Optics, Addison Wesley (Reading, MA) 1949.
198. Shakirova, Constructionof Petri Nets Via States of Action Objects And Subjects,
Proceedings of the 9th International Conference on Neural Information Processing (ICONIPOZ),
Vol. 1
199. Shandle, Who will Dominate the Desktop in the 90s?, Electronics, Feb 1990, pp 48-59.
200. Sharp, Networks and Interconnection Structures in An Introduction to Distributed and
Parallel Processing Computer Science Texts, Blackwell Scientific Publications, 1987
201. Sharples, Mike, et al, Computers and Thought, MIT Press (Cambridge) 1990.
202. Shubik, M., A Game Theoretic Approach to Political Economy, MIT Press (Cambridge,
MA), 1987.
203. Shubik, M., Game Theory in the Social Sciences, MIT Press (Cambridge, MA), 1984.
204. Silverman, Kaja, The Subject of Semiotics, Oxford University Press (New York), 1983.
205. Simon, H.A., The Sciences of the Artificial, MIT Press (Cambridge, MA), 1969.
206. Sirbu, M.A., D.P. Reed, An Optimal Investment Strategy Model for Fiber to the Home,
International Symposium on Subscriber Loops to the Home, pp. 149-155, 1988.
207. Sokal, R., P. Sneath, Principle of Numerical Taxonomy, Freeman (San Francisco) 1963.
208. Spector, Communication Support in Operating Systems for Distributed Transactions
Department of Computer Science, Carnegie-Mellon University, 1986
209. Speidel et al, Improved Hybrid Coders with 2-D Signal Processing for M oving Pictures
J. in Advances in Image Processing - Proceedings SPIE - The International Society for Optical
Engineering, 1987
210. Spulber, D.F., Regulation and Markets, MIT Press (Cambridge, MA), 1990.
211. Steiner, George, Heidegger, U Chicago (Chicago), 1978.
212. Steinmetz, Ralf, Synchronization Properties in Multimedia Systems, IEEE Journal on Sel
Areas in Comm, April, 1990, pp 401-412.
213. Strong, J., Concepts of Classical Optics, Freeman (San Francisco) 1958.
214. Sutcliffe, A., Human-Computer Interface Design, Springer Verlag (New York), 1988.
215. Swastek and Verreke, M igrating to FDDI on Your N ext Big LAN Installation in Data C
om m unications, June 21, 1989
216. Tannenbaum, A., Computer Communications, Prentice Hall (Englewood Cliffs, NJ),
1989.
355 | P a g e
217. Taylor, Mark C., Ceconstruction in Context, The university of Chicago Press (Chicago,
IL), 1986.
218. Tcha et al, Link-by-Link Bandwidth Allocation in an Integrated Voice/Data Network
Using the Fuzzy Set Approach in Computer Networks and ISDN System s, 16, Elsevier
Publishers, 1989
219. Temin, P., The Fall of the Bell System, Cambridge Univ Press (Cambridge), 1987.
220. Ten Hagen and Trienkens, Pattern Representation in Data Structures for Raster Graphics
edited by L.R.A. Kessener, FJ. Peters and M.L.P. van Lierop Eurographic Seminars, Springer-
Verlag, 1986
221. Teorey, Toby, James Fry, Design of Database Structures, Prentice Hall (Englewood
Cliffs, NJ) 1982.
222. Terry, D., D. Swinehart, Managing Stored Voice in the Etherphone System, ACM Trans
Cptr Sys, Vol 6, No 1, Feb 1988, pp3-27.
223. Teunissen, J. van den Bos, A Model for Raster Graphics Language Primitives edited by L
.R A . Kessener, F.J. Peters and M.L.P. van LieropEurographic Seminars, Springer-Verlag, 1986
224. Thomasian, A Perform ance Study of Dynam ic Load Balancing in Distributed System s
in 1987 IEEE 7th International Conference on Distributed Com puting System s edited by R.
Popescu-Zeletin, G. Le Lann, K.H. Kim Com puter Society Press of the IEEE, 1987
225. Thurber, Getting a Handle on FDDI in Data C om m unications, June 21, 1989
226. Tjaden, G., CATV and Voice Telecommunications, IEEE ICC, 1984. Toffler, A., The
Adaptive Corporation, Bantam (New York), 1985. Toy, S., Castles, Heinemann (London), 1939.
227. Tsichritzis, D. et al, A Multimedia Office Filing System, Proc VLDB 1983.
228. Tu et al, A Hybrid Coding Scheme for Still Images :, Adaptive Vector Quantization with
Correction Information in DCT Domain, Applications of Digital Image Processing X -
Proceedings, edited by A.G. Tescher, SPIE - The International Society for Optical Engineering,
1987
229. Ueltzen, Codex View s on LAN System s Integration Localnet 86, Online Publishers,
1986
230. Ullman, Jeffrey, Database and Knowledge Base Systems, Computer Science Press
(Rockville, MD) 1988.
231. van Lierop, Intermediate Data Structures for Display Algorithms, Data Structures for
Raster Graphics, edited by L.R.A. Kessener, F J. Peters and M.L.P. van Lierop, Eurographic
Seminars, Springer-Verlag, 1986
232. Vandermeulen et al, Surface Reconstruction from Planar Cross Sections for 3-D
Integrated Representation of Medical Images, in Advances in Image Processing - Proceedings
edited by A. Oosterlinck, A.G. Tescher SPIE - The International Society for Optical Engineering,
1987
233. Venetsanopoulos, Image Acquisition and Presentation, in Advances in Image Processing
and Pattern Recognition edited by V. Cappellini, R. Marconi North-Holland, 1986
356 | P a g e
234. Vickers, R., T. Vilmansen, The Evolution of Telecommunications Technology, Proc
IEEE, vol 74, No 9, pp 1231-1245, Sept 1986.
235. Von Auw, A., Heritage and Destiny, Praeger (New York), 1983.
236. Von Hippel, E. The Sources of Innovation, Oxford (New York), 1988.
237. Von Neumann, J., O. Morgenstern, Theory of Games and Economic Behavior, Wiley
(New York), 1944.
238. Walach et al, A Modified Block Truncation Coding Technique for Image Compression
inAdvances in Image Processing and Pattern Recognition, edited by V. Cappellini, R. Marconi
North-Holland, 1986
239. Warnke, Gadamer, Stanford (Stanford, CA), 1987.
240. Weinstein, Getting the Picture -A Guide to CATV and the New Electronic Media, IEEE
Press, 1986
241. Weizenbaum, Joseph, Computer Power and Human Reason, Freeman (New York) 1976.
242. Wendland, A Scanning Scheme for a New HDTV Standard, HDTV 85, Minister of
Supply and Services Canada, 1986
243. West, E.H., et al, Design, Operation, and Maintenance of a Multi Firm Shared Private
Network, IEEE MONECH Conf, pp80-82, 1987.
244. White, Image Transmission over Grade T elephone Lines in Applications o f D igital
Image Processing X - Proceedings edited by A.G. Tescher SPIE - The International Society for
Optical Engineering, 1987
245. Wiener, N., Cybernetics, MIT Press (Cambridge, MA), 1948.
246. Wiener, N., God and Golem, MIT Press (Cambridge, MA), 1964.
247. Wiener, N., The Human Use of Human Beings, Avon (New York), 1967.
248. Wilkes, M.V., The bandwidth Famine, Comm ACM, pp 19-21, Vol 33, No 8, Aug 1990.
249. Winograd, Terry, Fernando Flores, Understanding Computers and Cognition, Addison
Wesley (Reading, MA), 1987.
250. Wirth, T.E., Telecommunications in Transition, U.S. House of Representatives
Committee Report, Nov, 1981.
251. Wittgenstein, Ludwig, Philisophical Investigations, MacMillian Publishing Co. (New
York), 1958.
252. Wittgenstein, Ludwig, Tractatus, Routledge & Kegan Paul (London) 1922.
253. Woelk, D, W. Kim, Multimedia Information Management, VLDB Conf 1987.
254. Woelk, D. et al, An Object Oriented Approach to Multimedia Databases, ACM, 1986, pp
311-325.
255. Wolfe, T., From Bauhaus to Our House, Simon and Schuster (New York), 1981.
256. Wright, Karen, The Road to the Global Village, Scientific American, March 1990, pp 83-
94.
357 | P a g e
257. Zhgou and Ferrari, A Measurement Study of Load Balancing Performance 1987 IEEE
7th International Conference on Distributed C om puting System s Computer Society Press of the
IEEE, 1987
258. Zuboff, S., In the Age of the Smart Machine, Basic Books (New York), 1988.
259. Zurawski and Zhou, Petri Nets and Industrial Applications: A Tutorial, IEEE.
Transactions On Industrial Electronics, Vol. 41, No. 6, December 1994
358 | P a g e
13 INDEX
"throwness", 53 CATV, 12, 65, 186, 191, 196, 215, 218, 228,
abstraction, 24, 35, 39, 82, 83, 84, 153, 229, 230, 231, 232, 233, 242, 243, 244,
157, 179 356, 357
achromatic, 111, 114, 117, 118, 130 Central Office, 65, 196, 203, 225, 226
Action, 57, 63, 162, 353, 355 Centralized, 65, 97, 193, 200, 251, 292, 293
additive colors, 125 characterization, 35, 36, 39, 48, 96, 107,
Addressing, 249, 250, 258, 266 109, 110, 111, 132, 137, 146, 148, 151,
Alexandria, 14 153, 154, 157, 173, 174, 176, 177, 179,
ambiguity, 13, 56, 92, 105, 106, 156 290, 304
application, 23, 31, 37, 50, 63, 79, 93, 96, chroma, 121
97, 98, 103, 151, 154, 156, 157, 158, 161, CIE Chart, 123, 125
183, 212, 215, 218, 220, 241, 254, 256, CNRI, 70
285, 286, 290, 292, 294, 299, 301, 304, coding, 117, 118, 151, 253, 268
305, 307, 313, 324, 326, 327, 328, 329, color, 20, 42, 58, 111, 112, 113, 114, 118,
343 120, 121, 122, 124, 125, 130, 131, 132,
Arbib, 54, 346 133, 134, 135, 147, 150
architecture, 34, 36, 39, 64, 65, 67, 68, 69, command, 150, 264, 277, 303, 329, 337,
70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 340
92, 98, 104, 136, 146, 159, 160, 184, 185, communications, 10, 12, 13, 14, 15, 21, 22,
186, 187, 188, 190, 191, 192, 193, 194, 26, 27, 34, 35, 36, 38, 39, 41, 44, 45, 46,
195, 197, 198, 199, 200, 202, 203, 205, 47, 48, 49, 50, 51, 52, 54, 57, 58, 60, 61,
206, 207, 211, 212, 213, 215, 216, 224, 64, 65, 67, 68, 69, 70, 71, 73, 75, 77, 80,
225, 228, 237, 238, 241, 247, 249, 250, 81, 82, 83, 84, 86, 87, 89, 90, 92, 97, 98,
251, 269, 271, 275, 276, 277, 280, 282, 99, 101, 102, 103, 104, 105, 107, 110,
285, 292, 293, 294, 297, 299, 316, 321, 113, 133, 157, 158, 159, 162, 173, 175,
323, 331, 342 179, 181, 183, 184, 186, 190, 192, 196,
AT&T, 52, 64, 65, 193, 197, 219, 223, 226, 199, 205, 207, 210, 212, 213, 215, 216,
227, 231 218, 219, 220, 222, 223, 224, 225, 228,
bandwidth, 15, 31, 34, 42, 65, 69, 70, 76, 229, 236, 246, 247, 248, 249, 250, 251,
77, 84, 131, 135, 136, 137, 141, 190, 191, 252, 253, 254, 255, 256, 257, 258, 259,
192, 194, 195, 196, 208, 209, 210, 211, 261, 262, 263, 264, 265, 267, 268, 271,
212, 213, 226, 229, 233, 235, 249, 278, 273, 277, 278, 279, 280, 281, 282, 284,
296, 302, 337, 342, 357 285, 292, 293, 294, 297, 298, 299, 300,
bits, 20, 30, 34, 35, 36, 42, 59, 60, 65, 71, 301, 302, 303, 310, 312, 314, 315, 316,
84, 103, 107, 110, 116, 117, 118, 119, 323, 324, 325, 327, 328, 330, 336, 337,
134, 141, 142, 147, 151, 175, 177, 194, 343, 344, 345
208, 211, 225, 250, 252, 255, 256, 269, connection based, 34, 246, 255, 263, 266
270, 271, 280, 281, 304 Conversationality, 48, 57, 62
Breaking Down, 62 Cornea, 111, 112
bus network topology, 250 CPE, 65, 78, 192, 196, 197, 211, 212
canonical, 33, 78, 192, 325, 326, 328, 334, CRT, 133, 134, 135, 346, 354
335, 336, 338, 343 Database, 103, 300, 304, 310, 313, 344,
CAT, 20, 23, 24, 35, 37, 128, 301, 304 347, 349, 351, 356
359 | P a g e
DBDQ, 273 289, 297, 298, 300, 302, 303, 304, 310,
decomposability, 60 315, 316, 321, 324, 326, 334, 340, 343
Deconstructionism, 49 entertainment, 15, 186, 213, 214, 215, 218,
Delbruck, 42, 45, 348 229, 231
Dertouzos, 64, 69, 70, 193, 348 entropy, 13
Device, 324, 334, 339, 340 epistemology, 41, 343
Device management, 324, 334, 339, 340 eye, 20, 110, 111, 112, 113, 114, 117, 118,
diffuse, 113 120, 122, 130, 131, 132, 134, 135, 196
Discrimination, 113, 114 FDDI, 76, 190, 269, 270, 271, 272, 273,
Disk, 180, 336, 348 280, 349, 351, 355, 356
Displaceable, 46 Felde, 15
display, 15, 18, 19, 20, 21, 27, 30, 31, 32, field of view, 135
34, 38, 41, 45, 46, 47, 49, 50, 51, 55, 60, File, 33, 161, 180, 181, 260, 328, 333, 334,
110, 111, 116, 117, 133, 134, 135, 136, 335, 352
146, 147, 148, 149, 150, 151, 152, 158, File Management, 335
159, 160, 161, 169, 180, 212, 263, 275, film, 14, 38, 44, 107, 111, 113, 115, 116,
312, 313, 334 117
distributed, 12, 34, 39, 47, 59, 61, 64, 65, formats, 13, 19, 21, 68, 70, 107, 122, 180,
71, 73, 77, 92, 93, 97, 98, 105, 112, 174, 207
185, 187, 192, 193, 194, 196, 197, 201, Fourier Transform, 127, 128, 129, 130
202, 205, 208, 209, 211, 213, 214, 229, frequencies, 35, 114, 116, 120, 139
230, 233, 238, 241, 256, 262, 265, 268, Gadamer, 53, 54, 55, 69, 88, 89, 90, 91, 190,
272, 279, 285, 286, 291, 293, 294, 298, 349, 357
299, 300, 301, 302,303, 311, 312, 313, Gbps, 31, 36, 69, 154, 183, 195, 207, 246,
315, 316, 317, 318, 321, 323, 324, 325, 248, 268, 269
328, 331, 333, 335, 336, 337, 338, 339, GKS, 32, 146, 148, 149, 150, 151, 152, 161
340, 341 grating, 128
Distributed, 98, 201, 211, 251, 269, 271, gray scale, 20, 116, 117, 118, 119, 120
293, 299, 300, 310, 313, 321, 323, 324, Gutenberg, 14, 353
331, 339, 341, 344, 347, 349, 352, 353, halftone, 118, 119, 120
355, 356, 358 HDTV, 50, 113, 357
Distributed Operating System, 331, 344 Heidegger, 10, 53, 54, 55, 59, 74, 85, 86,
DOS, 161, 324, 325, 329, 331, 336, 338 188, 350, 355
Drucker, 14, 43, 44, 75, 189, 348 hermeneutic, 75, 87, 88, 89, 90, 91, 92, 190
ear, 137, 138, 139 Hermeneutics, 49, 53, 87, 88, 89, 92, 349,
educated, 15 352
element, 13, 15, 20, 21, 22, 23, 24, 35, 36, hue, 121
39, 42, 45, 46, 47, 48, 49, 52, 58, 59, 62, Hyper-Dimensionality, 48
63, 65, 66, 75, 78, 80, 86, 93, 96, 99, 100, hypermedia, 48, 49
102, 104, 107, 110, 112, 114, 121, 149, I/O, 32, 158, 248, 301, 302, 311, 312, 323,
154, 157, 161, 162, 164, 165, 171, 177, 324, 329, 333, 334, 335, 337, 338, 340
179, 180, 181, 187, 189, 190, 193, 194, I/O Management, 324, 334, 340
195, 196,197, 201, 203, 210, 211, 213, image, 10, 12, 13, 14, 15, 16, 19, 20, 21, 23,
215, 246, 249, 251, 252, 256, 257, 261, 24, 31, 34, 35, 36, 38, 39, 42, 45, 46, 47,
262, 267, 271, 275, 277, 278, 282, 286, 53, 58, 59, 60, 70, 97, 102, 103, 104, 107,
109, 110, 111, 112, 113, 114, 116, 117,
360 | P a g e
118, 119, 131, 133, 134, 135, 136, 145, Local Operating System, 329
147, 150, 151, 152, 154, 157, 159, 173, Localized, 251
176, 179, 212, 213, 246, 247, 256, 261, LOS, 325, 329, 338
262, 265, 281, 282, 291, 301, 304, 306, Luminosity, 133
319, 320 MAC, 18, 21, 72, 146, 256, 268, 270, 271
inband, 251 Mac Cormac, 54, 352
infrastructure, 63, 64, 66, 67, 68, 69, 70, 71, Mbps, 34, 36, 76, 97, 183, 190, 195, 209,
72, 73, 80, 183, 184, 185, 186, 198, 205, 226, 229, 235, 238, 246, 247, 268, 269,
206, 207, 209, 212, 215, 216, 220, 221, 270, 271, 291, 333
223, 224, 225, 228, 231, 236, 240, 241, McLuhan, 14, 21, 43, 44, 50, 75, 76, 86,
242, 248, 265 188, 189, 345, 353
Interactive, 32, 46, 146, 348, 354 media, 12, 13, 14, 15, 16, 21, 37, 38, 39, 41,
INTERNET, 68, 207 42, 44, 45, 46, 48, 51, 54, 55, 64, 72, 75,
interpreter, 91, 329 90, 105, 107, 108, 110, 111, 146, 152,
Iris, 111, 112 154, 180, 189, 213, 247, 279, 303, 304,
ISDN, 77, 191, 195, 227, 247, 251, 256, 319, 322, 325, 334, 341, 343, 345
269, 273, 347, 349, 352, 353, 356 Media Distributed Operating System,
ISO, 29, 68, 98, 207, 262, 297, 323, 327, 324, 331
341, 351 MEDOS, 324, 325, 331, 333
Kahn, 70, 77, 192, 194, 195, 196, 197, 198, memory, 16, 34, 90, 107, 110, 161, 180,
223, 227, 351 301, 302, 310, 315, 323, 326, 329, 334,
Kbps, 35, 59, 76, 77, 191, 192, 195, 208, 336, 337, 338, 340, 341
209, 247, 257 Memory, 179, 180, 315, 324, 333, 334, 335,
kernel, 34, 150, 276, 326, 331, 334 340
Kernel Architecture, 332 Memory Management, 324, 333, 334, 340
knowledge, 14, 16, 21, 35, 41, 43, 44, 55, message, 12, 13, 23, 24, 42, 43, 44, 48, 49,
57, 70, 76, 79, 82, 85, 87, 89, 99, 105, 50, 64, 76, 86, 87, 90, 95, 97, 98, 175,
183, 189, 214, 343, 345 176, 189, 211, 251, 256, 257, 258, 261,
Kuhn, 55, 74, 75, 76, 78, 188, 189, 206, 351 265, 267, 270, 272, 288, 291, 294, 296
Language, 20, 32, 56, 57, 90, 155, 344, 353, messenger, 48, 49, 50, 64, 87, 248
356 MIP, 311, 338
layer, 31, 32, 34, 36, 68, 92, 93, 95, 99, 105, MMDB, 321
111, 194, 207, 237, 247, 250, 252, 253, Morrill, 42, 353
254, 255, 256, 257, 258, 259, 260, 261, MRI, 20, 24, 35, 37, 82, 301, 304, 305, 320
262, 263, 264, 265, 266, 267, 268, 273, multimedia, 10, 12, 13, 14, 15, 16, 18, 19,
274, 276, 280, 283, 284, 285, 286, 289, 20, 21, 22, 23, 24, 26, 27, 31, 34, 35, 36,
295, 297, 298, 299, 316, 323, 326, 327, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
328, 329, 330, 334 48, 49, 50, 51, 52, 53, 54, 56, 57, 58, 59,
Layer 1 Physical, 256 60, 61, 62, 63, 67, 72, 73, 80, 81, 82, 83,
Layer 2 Data Link, 255 84, 85, 86, 87, 88, 89, 91, 92, 93, 96, 98,
Layer 3 Network, 255 102, 103, 104, 105, 106, 107, 113, 114,
Layer 4 Transport, 255 133, 137, 143, 153, 154, 155, 157, 158,
Layer 5 Session, 36, 254 159, 173, 177, 179, 180, 181, 212, 213,
Layer 6 Presentation, 254 246, 247, 248, 249, 250, 251, 252, 253,
Lens, 112 254, 255, 256, 257, 261, 262, 263, 264,
lightness, 121 266, 269, 273, 274, 278, 279, 280, 281,
361 | P a g e
282, 283, 285, 290, 297, 298, 299, 302, 267, 270, 273, 274, 278, 279, 280, 286,
303, 304, 305, 306, 310, 312, 319, 320, 297, 299, 301, 314, 323, 337, 338, 341,
321, 323, 324, 331, 333, 334, 341, 343, 343, 345
344, 345 Petri Net, 164, 165, 166, 169, 171, 174,
Multiple processes per Multiple 348, 354
Processors (MP), 339 PHIGS, 32, 146, 151, 152, 161, 354
Multiple Processes per philosophy, 10, 43, 46, 50, 85, 233, 343
Processor/Multiple processors, 339 photograph, 24, 110
Multi-Sensory, 46 Photosensitivity, 113
multiuser, 13, 14, 21, 33, 158, 299, 302, physical, 23, 26, 35, 38, 39, 49, 58, 65, 66,
343 67, 68, 69, 71, 72, 73, 80, 84, 89, 100,
Multi-User, 46 101, 102, 110, 158, 184, 185, 194, 203,
network management, 65, 93, 98, 99, 193, 206, 207, 212, 213, 217, 240, 241, 242,
213, 237, 267, 273, 274, 275, 277, 278, 250, 252, 255, 256, 257, 258, 266, 268,
286, 297, 298, 323 273, 274, 280, 300, 310, 321, 334, 336,
Network Operating System, 330 339, 340
Newton, 14, 125 pixel, 20, 30, 36, 60, 116, 117, 118, 133,
nm, 112, 122, 131 134, 147, 151
NREN, 68, 69, 76, 183, 184, 185, 190, 195, PostScript, 146, 152, 161
206, 207, 220, 228, 238, 239, 242, 351 presentation, 13, 14, 19, 39, 42, 43, 45, 48,
NTSC, 135, 136, 263 50, 56, 75, 97, 105, 120, 154, 156, 157,
Nyquist, 141 158, 162, 163, 165, 189, 242, 254, 256,
object, 24, 36, 46, 47, 48, 58, 59, 60, 62, 63, 260, 261, 262, 263, 270, 280, 291, 299,
80, 81, 92, 93, 96, 97, 98, 100, 101, 103, 328, 341, 345
104, 133, 152, 256, 257, 261, 262, 280, printed, 13, 79, 86
281, 282, 285, 290, 292, 294, 296, 304, Process, 47, 162, 324, 333, 334, 335, 338,
320 339, 348
operating system, 31, 39, 98, 158, 160, 161, Process Management, 324, 338
213, 260, 293, 300, 323, 324, 325, 326, psychological, 117, 135, 343
327, 328, 329, 330, 331, 333, 334, 335, psychometric, 131, 132, 133
336, 338, 339, 340, 341 Pupil, 112
Optic Nerve, 112 quantization, 109, 117, 118, 120, 141, 142
OS, 324, 327, 329, 330, 333, 334, 335, 347 QWERTY, 155
Packet, 34, 66, 196 RAM, 180, 336, 340
paradigm, 21, 42, 43, 53, 59, 74, 75, 76, 78, rate, 31, 35, 36, 60, 70, 76, 77, 103, 110,
79, 80, 98, 105, 107, 108, 152, 154, 189, 135, 136, 141, 157, 173, 177, 183, 184,
195, 210, 216, 238, 241, 242, 279, 280, 190, 191, 192, 195, 208, 209, 210, 225,
294, 302, 308 226, 227, 241, 246, 269, 270, 271, 282,
paradigms, 55, 62, 73, 74, 75, 79, 80, 105, 296, 337
186, 187, 188, 189, 199, 206, 216, 241, regeneration, 143, 229
323, 344 resolution, 14, 18, 34, 35, 36, 50, 64, 81,
Paradigms, 73, 74, 78, 343 82, 111, 112, 113, 116, 117, 134, 152,
performance, 39, 55, 66, 93, 99, 107, 154, 154, 159, 193
157, 173, 176, 179, 180, 181, 197, 208, retina, 42, 111, 112, 113, 114, 117, 118,
209, 212, 213, 216, 223, 225, 226, 227, 130
229, 230, 238, 248, 249, 252, 263, 266, Retina, 112
362 | P a g e
RGB, 122, 123, 125, 133, 134, 135 switches, 65, 76, 190, 196, 211, 212, 227,
ring topology, 250 235, 247
Semiotics, 49, 99, 100, 101, 102, 348, 355 synch, 98, 294, 296, 297
services, 33, 34, 36, 51, 52, 61, 62, 63, 67, Synchronization, 61, 93, 96, 103, 169, 263,
69, 74, 92, 93, 95, 98, 104, 162, 183, 185, 264, 282, 285, 290, 291, 292, 294, 296,
188, 202, 204, 209, 213, 214, 215, 218, 340, 352, 355
219, 220, 222, 223, 225, 231, 234, 235, taste, 45, 107, 110
238, 242, 246, 248, 254, 256, 258, 259, telegraphy, 13
260, 261, 262, 263, 264, 266, 267, 271, television, 44, 81, 113, 135, 136, 215, 228,
274, 280, 284,285, 286, 289, 294, 298, 242, 249
299, 327, 328, 329, 330, 334, 345 Throwness, 62, 74, 188
session, 13, 23, 26, 33, 34, 36, 39, 56, 57, touch, 21, 36, 45, 50, 81, 107, 110, 156, 279
58, 61, 63, 92, 93, 94, 95, 96, 97, 98, 99, Traffic, 340
105, 106, 162, 173, 179, 246, 247, 255, transfer, 13, 14, 15, 31, 46, 48, 54, 110,
256, 257, 259, 260, 261, 262, 263, 264, 152, 180, 213, 216, 223, 227, 240, 241,
265, 266, 267, 280, 283, 284, 285, 286, 256, 269, 270, 272, 273, 312, 337
287, 288, 289, 290, 291, 292, 293, 294, transmission, 12, 13, 14, 36, 76, 77, 82, 88,
295, 296, 297,298, 299, 300, 301, 302, 90, 94, 97, 99, 102, 127, 137, 138, 190,
312, 323, 324, 328, 343 192, 194, 195, 200, 201, 209, 210, 228,
Shandle, 43, 355 231, 232, 237, 255, 257, 266, 267, 271,
Shell Architecture, 332 287, 291, 292, 298, 343
sight, 21, 45, 50, 57, 107, 108, 279 Transparency, 62
Single process per Multiple tree topology, 249
processor/MP, 339 UNIX, 157, 158, 161, 219, 280, 298, 324,
Single process per Processor/Multiple 327, 328, 336
Processors, 339 Value, 51, 63, 216, 218, 221
sizing, 39, 107, 131, 154, 169, 173, 179, video, 13, 14, 19, 23, 36, 38, 41, 42, 43, 45,
181, 249, 278, 323 46, 47, 49, 59, 60, 77, 81, 84, 96, 103,
SMDS, 213, 230, 269, 271, 272, 273, 280 110, 135, 136, 145, 146, 172, 180, 181,
smell, 45, 81, 107, 110 192, 212, 213, 214, 215, 229, 230, 242,
software, 31, 43, 46, 50, 51, 66, 77, 146, 243, 244, 247, 255, 256, 261, 262, 263,
192, 196, 213, 246, 251, 252, 253, 266, 275, 280, 281, 282, 290, 304, 319, 320
275, 276, 323, 324, 329 virtual, 23, 26, 33, 36, 58, 92, 94, 223, 247,
sound, 19, 20, 21, 45, 50, 57, 81, 107, 108, 254, 263, 266, 285, 286, 334, 337
137, 138, 139, 279 visible spectrum, 121, 127
spectrometer, 127, 128 voice, 12, 13, 15, 19, 21, 22, 23, 35, 43, 45,
star topologies, 250 46, 47, 52, 54, 56, 59, 60, 64, 65, 66, 75,
storage, 13, 16, 31, 39, 41, 42, 43, 45, 46, 76, 77, 102, 103, 110, 136, 137, 139, 140,
47, 51, 55, 71, 102, 107, 108, 110, 134, 141, 145, 146, 180, 181, 186, 190, 191,
135, 136, 137, 143, 146, 150, 151, 179, 192, 193, 195, 196, 197, 201, 202, 208,
180, 181, 213, 253, 279, 280, 301, 303, 209, 210, 212, 213, 224, 225, 226, 228,
310, 311, 314, 315, 320, 321, 331, 334, 231, 233, 236, 247, 251, 256, 261, 262,
340, 343 280, 281, 282, 304, 306, 319, 320
subsession, 94 wavelength, 122, 123, 127, 130, 131, 132
subsessions, 94 Winograd and Flores, 43, 53, 55, 56, 59, 74,
subtraction of light, 126 84, 85, 91, 188, 190
363 | P a g e
world view, 21, 49, 55, 64, 66, 67, 68, 69, 193, 195, 202, 203, 205, 206, 207, 208,
71, 74, 75, 76, 77, 78, 79, 80, 105, 108, 210, 211, 212, 223, 227, 241
184, 186, 187, 188, 189, 190, 191, 192, X-Ray, 24
364 | P a g e
View publication stats

Multimedia Comm03

Uploaded by

Copyright:

Available Formats

Multimedia Comm03

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Multimedia Comm03

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Multimedia Communications (Revised)

Book · October 2020

Terrence Patrick McGarty

The user has requested enhancement of the downloaded file.

Terrence P. McGarty Ph.D.

There is a philosophy of multimedia communications, a philosophy as to what we are trying to

Multimedia Communications must be "displaced human interactions". It is a simple concept but

Terrence P. McGarty Waterville Valley, NH September, 1989

1.1 DEFINITION OF MULTIMEDIA

1.1.1 The Image

1.1.2 The User

1.2 CURRENT SYSTEMS

Figure: Sample Current Systems

Figure: Sample Multimedia Computers

1.2.2 Presentation Formats

Figure: Sample Integrated Presentation Formats (Muriel Cooper's)

1.2.3 Interaction Formats

1.3 THE PARADIGM

The development of a theory for multimedia communications requires the development of a

Figure: The Multimedia Paradigm

1.3.2 Physical User

Figure : Multimedia Elementals

Figure: Development of Sessions and Subsessions

Figure: Multimedia Layering and Interfaces

Link Network Transport Session Link

1.4.2 Applications Software

4. I/O Programs (IGES, COM)

Figure: Software Architecture - Software Architecture for a Multimedia Messaging System.

1.4.4 1.4.4 Network Software

1.4.5 1.4.5 Storage

1.5 KEY PROBLEMS

1.5.1 Problem 1: Characterization

1.5.2 Problem 2: Utilization

1.6 PROCESSOR/ PACKETIZER

Figure: Mixing Concept in Printing Environment

addresses the issues of communications. We focus on the standard approach to communications

2.1 CURRENT DEFINITIONS OF MULTIMEDIA

Multimedia concepts encompass a wide variety of interpretations. To some, the multimedia

We attempt to develop a theory of multimedia communications, that encompasses the elements

1. A transition from an olfactory and tactile creature to a visual creature.

2.2 PHILOSOPHICAL PILLARS

To quote Drucker, who paraphrased McLuhan;

1.3 Multimedia Structures

Multimedia communications is characterized by the following factors:

1. Multi-Sensory: It uses several of the human senses in transferring, processing, and

2. Multi-User: It interconnects several users of the information into a conversational mode

3. Displaceable: It allows for the establishment of communications and information transfer

7. Communicate: Communications means the development of a conversational mode.

2. Communication and Conversationality: This characteristic is one of allowing for a

4. Presentation and Interaction: The interactiveness of the environment is a key element of

5. Non-Linearity and Hyper-Dimensionality: This dimension of characterization allows for

6. Sense-Complexity and Representation: This dimension allows for the combining of

3. Messenger: The messenger generally is thought of as an individual. In our context, the

2.3 MULTIMEDIA; ELEMENTS AND STRUCTURE

2.3.1 Elements of Multimedia

1. Displays: The MM display areas focuses on enhancement of images, improved

2.3.2 Value Creation in Multimedia

2.3.4 Elements and Structures

Our implicit beliefs and assumptions cannot be made explicit.