The Structure of

Digital Computing
From Mainframes to Big Data

by

Robert L. Grossman

Open Data Press LLC

2012
 c 2012 by Robert L. Grossman
Copyright �

Published by Open Data Press LLC

400 Lathrop Ave, Suite 90
River Forest, IL 60305, USA

All rights reserved.

Printed in the United States of America.

First Printing: June, 2012

Library of Congress Control Number: 2012908445

ISBN 978-1-936298-00-6
Contents

1 The Five Eras of Computing 1

1.1 Introduction . . . . . . . . . . . . . . . . . . 1
1.2 The Main Themes . . . . . . . . . . . . . . 2
1.3 A Billion IP Addresses . . . . . . . . . . . . 5
1.4 The SAD History of Computing . . . . . . 9
1.5 Why Symbols Matter . . . . . . . . . . . . 11
1.6 Algorithms as Recipes . . . . . . . . . . . . 15
1.7 Computing Devices . . . . . . . . . . . . . . 17
1.8 Case Study: Slide Rule . . . . . . . . . . . 19
1.9 From Mainframes to Devices . . . . . . . . 23
1.10 Mainframe Era . . . . . . . . . . . . . . . . 26
1.11 Case Study: Punch Cards . . . . . . . . . . 27
1.12 Personal Computer Era . . . . . . . . . . . 28
1.13 The Web Era . . . . . . . . . . . . . . . . . 30
1.14 Case Study: SMTP . . . . . . . . . . . . . . 31
1.15 Clouds of Devices . . . . . . . . . . . . . . . 34
1.16 Case Study: Routers . . . . . . . . . . . . . 36
1.17 The First Half Century of Computing . . . 38
1.18 The Commoditization of Data . . . . . . . . 43

2 Commoditization 45
2.1 Christmas and Easter . . . . . . . . . . . . 45
2.2 The Commoditization of Time . . . . . . . 47
2.3 The Commoditization of Space . . . . . . . 49
2.4 Moore’s Law . . . . . . . . . . . . . . . . . 52
2.5 Commoditization is All Around Us . . . . 54
2.6 The Doubling Game . . . . . . . . . . . . . 58

i
 2.7 Transforming Technologies . . . . . . . . . 61
2.8 Storage and Johnson’s Law . . . . . . . . . 62
2.9 Bandwidth and Gilder’s Law . . . . . . . . 66
2.10 Software and Stallman’s Law . . . . . . . . 67
2.11 Data and the Bermuda Principles . . . . . 74
2.12 Network Effects . . . . . . . . . . . . . . . 75

3 Technical Innovation vs. Market Clutter 81

3.1 Innovation vs. Clutter . . . . . . . . . . . . 81
3.2 Approximating Solutions to Equations . . . 83
3.3 Case Study: Business Intelligence . . . . . . 87
3.4 Views of Technical Innovation . . . . . . . . 88
3.5 The Imperative to be in the Upper Right . 91
3.6 Why Clutter Is Inevitable . . . . . . . . . . 93
3.7 Who Clutters . . . . . . . . . . . . . . . . 96
3.8 Sources of Clutter: Features . . . . . . . . 98
3.9 Case Study: Databases . . . . . . . . . . . . 101
3.10 Case Study: Searching for Primes . . . . . . 109
3.11 Case Study: Routing Packets . . . . . . . . 111

4 Technology Adoption Cycles 119

4.1 Forces Effecting Technology Adoption . . . 119
4.2 The Basic Equation of Marketing . . . . . 122
4.3 Getting to Main Street . . . . . . . . . . . . 127
4.4 Case Study: The Nike Pegasus . . . . . . . 130
4.5 Technology Roadmaps . . . . . . . . . . . . 133
4.6 Case Study: Clusters . . . . . . . . . . . . . 137
4.7 Context . . . . . . . . . . . . . . . . . . . . 140
4.8 Case Study: Relational Databases . . . . . 145
4.9 Technology Pain Points . . . . . . . . . . . 149
4.10 Case Study: Adoption of Linux . . . . . . . 156

5 The Era of Data 161

5.1 Introduction . . . . . . . . . . . . . . . . . . 161
5.2 Thinking about Big Data . . . . . . . . . . 162
5.3 The Commoditization of Data . . . . . . . 164
5.4 The Data Gap . . . . . . . . . . . . . . . . 168
5.5 Extracting Knowledge from Data . . . . . 172

ii
 5.6 Kepler’s Law and Brahe’s Data . . . . . . . 178
5.7 Pearson’s Law . . . . . . . . . . . . . . . . . 182
5.8 The Bermuda Principles . . . . . . . . . . . 188
5.9 World Population . . . . . . . . . . . . . . . 190
5.10 The Shape of Data . . . . . . . . . . . . . 193
5.11 Case Study: Consumer Databases . . . . . 199
5.12 Creating Digital Data . . . . . . . . . . . . 203
5.13 Using Data to Make Decisions . . . . . . . . 207
5.14 Case Study: Mammograms . . . . . . . . . 211
5.15 Events, Profiles and Alerts . . . . . . . . . . 214
5.16 Case Study: NASA’s EOS . . . . . . . . . 219

Notes 225

References 249

iii
Preface

This book is about the structure of digital computing: what

is significant, what is novel, what endures, and why it is all
so confusing. The book tries to balance two point of views:
digital computing as viewed from a business perspective,
where the focus is on marketing and selling, and digital
computing from a more technical perspective, where the
focus is on developing new technology.
My goal was to write a short book about digital com-
puting that takes a long term point of view and integrates
to some extent these two perspectives.
The book is shaped by my personal experience in these
two worlds: From 1996–2001, I was the Founder and the
CEO of a company called Magnify, Inc. that developed and
marketed software for managing and analyzing big data.
Prior to this, from 1988–1996, I was faculty member at the
University of Illinois at Chicago (UIC), where I did research
on data intensive and distributed computing. From 1996–
2010, I remained at UIC as a part time faculty member.
I wrote the sections in this book over an approximately
eight year period from 2001 to 2008, with most of the writ-
ing done during 2001–2003. I have left the older sections
by and large as they were originally written.

v
Although there have been some changes since 2003 (for ex-
ample, computers are faster, there are more web sites, and
phones are smarter), hopefully as the book will make clear,
at a more fundamental level, we are still on the same fifty
or so year trajectory today that we were on in 2003.

Robert L. Grossman

vi
Chapter 1

The Five Eras of

Computing

1.1 Introduction
This book is about the structure of digital computing: it
is concerned with what is significant, what is novel, what
endures, and why it is all so confusing.
Computing and communication technologies have got-
ten a bad name for being hard to predict and difficult to
understand. In this book, I take the opposing point of view:
that many of the most important phenomena that underlie
computing have been remarkably regular and predictable
over the past fifty years.
For example, the remarkable growth of processing power
exemplified by Moore’s Law has followed a regular pattern
for over forty years. To put it simply, for most applica-
tions, processing power is a commodity and no harder to
get than other commodities, such as electrical power. What
is sometimes not appreciated is that a variety of other un-
derlying processes that form the basis for today’s compu-
tational and communications infrastructure have also been
commoditized. For example, software and network band-
width have been commoditized and show a similar regular-
ity.

1
2 The Five Eras of Computing

On the other hand, it is easy to lose sight of this regular-

ity and predictability given the market clutter created by
the many players with financial interests in computing and
related fields. One of themes of this book is that technical
innovation is generally masked by market clutter.
Understanding technology is often confused with the
challenge of predicting which of the several thousand tech-
nology vendors will be around in five years and what their
sales and profitability will be. This is a much harder prob-
lem and not one of the subjects of this book. It may be
helpful to think of the survival of a vendor over five years
as being modeled by a random walk, in much the same way
that the stock market is often modeled by a random walk.
True computing innovations have a beauty and a longe-
vity that creates regularity and simplicity in the historical
narrative of computing. Although technical innovations are
rare and cannot be predicted, they are usually recognized
relatively quickly.
In this book, we try to focus on some of the underlying
ideas and principles which have been fundamental drivers
for computing and communications technology. These tend
to be simple, rather than complex; long-lived, rather than
short-lived; and easy to understand, but not easy to antic-
ipate. These drivers have broad applicability rather than
narrow applicability. Although they may have been intro-
duced by individuals, they are brought to market by a va-
riety of vendors using a variety of business models over a
number of business cycles.

1.2 The Main Themes

Everything should be made as simple as possible, but no
simpler.
Albert Einstein

This book has five main themes:

1.2 The Main Themes 3

Theme 1. Our computing environment is shaped

by commoditization, which governs the progress
from one computing era to the next. The most
familiar example of commoditization is Moore’s law de-
scribing the rapid increase in power of integrated circuits
at the same time that the unit cost has been stable or
decreased. More generally, commoditization is the phe-
nomenon in which unit capacity of a core technology grows
exponentially, while unit cost is stable or decreases.
In this book we also examine the commoditization of
several other critical technologies including storage, net-
working, and software. One of the themes in this book is
that the commoditization of a handful of critical technolo-
gies has shaped our computing and communications infras-
tructure. We call these transforming technologies. Most
other computing related technologies merely fill in the de-
tails.
The process of adopting and using transforming tech-
nologies is regular, lasts for decades, and is relatively easy
to forecast. It is also not new. The printing press commodi-
tized books and the telephone commoditized global person
to person communication.
Chapter 1 provides an introduction to five different eras
of computing, each one shaped by the commoditization of
a different component of technology. Chapter 2 discusses
commoditization in more detail.

Theme 2. Technical innovation is rare. It is also

difficult to predict. Although technical innovation is im-
portant and critical, given its rarity, there is ample time to
detect it and to understand it when it does occur.
You can think of innovation as being at the bottom of an
inverted pyramid. Look it this way: A technical innovation
requires at least 10 engineering advances. A hundred com-
panies will try to bring products to market commercializing
these advances. These companies over the lifetime of the
products will produce 1000 marketing campaigns. Analysts
and pundits will write 10,000 articles analyzing these cam-
4 The Five Eras of Computing

paigns and products. It is more efficient to understand the

single innovation than to deconstruct the 10,000 articles.
Theme 3. Market clutter is rampant. The large
number of players with financial stakes in technology pro-
duces a large amount of market clutter. It is hard to see
through the market clutter. Pundits and industry analysts
are part of the system and instead of helping, they make
things worse. There is not much to do about this except to
ignore it.
As already mentioned, in this book we tend to view the
survival of a technology company over a five year period as
a random walk: looking at five year intervals, some technol-
ogy vendors will grow, some will shrink, and some will be
absorbed by other vendors; but to first order, which vendor
does which can best be modeled as random walk.
Chapter 3 is about the rarity of technical innovation
and the confusion caused by market clutter.
Theme 4. Technology takes time. The process by
which technology is created and adopted takes years. The
process begins in the laboratory and ends with sales and
marketing. This process is relatively well understood and
relatively regular. On the other hand, the particular ven-
dors that bring a new technology to market are not so easy
to predict. Which vendors survive and which fail is perhaps
best viewed as a random walk, as we have mentioned. The
adoption of new technology is complicated by several differ-
ent cycles involved: technology adoption cycles usually last
a decade or more, while the life cycle of many technology
companies is 3-7 years, and marketing cycles and fashions
last 1 to 2 years. You can think of this as a tricycle with
three different size wheels. Steering is obviously quite hard.

Chapter 4 is about technology adoption cycles.

Theme 5. We are entering the era of big data. We
are currently entering an era defined by the commoditiza-
tion of data. More precisely, over the next decade or so,
we will be entering a new era of discovery driven by vast
1.3 A Billion IP Addresses 5

amounts of new data being produced and archived. From

a broad historical perspective, having a surplus of unana-
lyzed data is unusual. For example, Darwin spent 17 years
collecting data before publishing one of his papers. On the
other hand, today an individual with a laptop and a web
connection can try to create new drugs by accessing human
genetic sequences, three dimensional protein data, and the
chemical properties of various compounds. All this is avail-
able from the web today at no cost.
Chapter 5 is about this emerging era of data.
How the book is organized. The chapters are meant
to be read in order. As just mentioned, Chapter 1 pro-
vides a framework for understanding the most important
broad trends in computing. Chapter 2 is about commoditi-
zation. Chapter 3 is about the rarity of innovation and the
prevalence of market clutter. It contains lots of examples,
the purpose of which is to give the reader some practice so
that it is easier to separate technical innovation from mar-
ket clutter. Chapter 4 describes the pattern with which
new technology is typically adopted by the market and the
many years this usually requires. Chapter 5 is about the
Era of Big Data.
Each chapter contains several extended examples and
case studies, some of which are quite detailed. Feel free to
skim or to ignore the ones that you don’t find interesting.
These examples and case studies are included since many
people find it easiest to learn through concrete examples.

1.3 Case Study: A Billion IP Addresses

When you use your home computer to buy an airline ticket
on the web site www.united.com, your computer looks up
the web site www.united.com to get the number 209.87.-
112.90. This is called an IP address and in some ways
is somewhat similar to a telephone number. The current
format for IP addresses was defined in 1981 and provided
roughly 4 billion of them. Since the 1981 world population
6 The Five Eras of Computing

9 44 40 40 209.247.34.166 internap-ne.chicago1.level3.net
10 38 37 37 64.94.32.11 border6.po1-bbnet1.chg.pnap.net
11 49 40 41 64.94.34.74 mypoints10.border6.chg.pnap.net
12 45 40 40 209.87.127.111 -
13 42 40 43 209.87.112.90 www.united.com

Figure 1.1: The Linux command traceroute provides the

IP addresses of intermediate points between your computer
and hosts on the Internet, such as www.united.com. This
is fragment of a traceroute to www.united.com, showing
the last portion of the route to www.united.com. The fifth
column is the IP address of the intermediate points along
the way to www.united.com.

was roughly 4.5 billion, since only a handful of people had

access to computers, and since only some of these had net-
work access, this seem a reasonable number of addresses.
To connect to the Internet, a company such as United
Airlines needs an IP address such as 209.87.112.90. Once it
has an IP address, it can provide a variety of services, such
as serving web pages describing flights between Chicago
and Hawaii and offering airline tickets for sale.
Beginning in 1999, a new type of Internet address be-
came available, called IPv6. An example of an IPv6 address
is
1080:0:0:0:8:800:200C:417A.
IPv6 addresses are longer than IPv4 addresses. IPv4 ad-
dresses are 32 bits long, while IPv6 addresses are 128 bits
long.
Today, not only can computers connect to the Internet,
but so can mobile phones. This means that it is useful for
a mobile phone to have an IP number. For over a decade it
has been clear that there were not enough IPv4 addresses
for each device, such as a mobile telephone, to have its own
IPv4 number. The IPv6 addresses were introduced in part
so that each device could have its own IP number and easily
connect to the Internet.
1.3 A Billion IP Addresses 7

A manufacturer of mobile telephones, such as Nokia

or Ericsson, is assigned large blocks of IPv6 addresses to
burn in to the telephones they build. Blocks are assigned
in units called /48’s. For example, Nokia would request a
/48 from European Registry for the delegation of Internet
Numbers or ERIN. The interesting thing is that a /48 has
enough IP numbers to set up 216 separate networks, with
each network having as many as 264 separate computers or
other IP devices [62].
Since the world population is about 6 billion (or about
30
2 , this may seem somewhat excessive. On the other hand,
there were about 1.18 billion new telephones sold in 2008,
about 39% of the them by Nokia [109]. Today, we are in
the midst of a transition from a computing infrastructure in
which computers are connected to form networks that are
in turn aggregated to form the Internet, to an infrastructure
in which mobile devices supplying services are connected to
form networks that are in turn connected to create clouds of
services. A Nokia phone may require several IP addresses,
each for a separate service in the cloud, such as talking,
browsing the web, GPS location, etc. From this perspec-
tive, a billion addresses doesn’t go as far as once did.
The Internet interconnects millions of different networks
and billions of different computers and devices. Each com-
puter which is directly on the Internet has a unique IP
address. The simplest of are obtained by concatenating a
network ID and a host ID. The network ID identifies which
network it is on, while the host ID distinguishes different
computers on the same network.
Actually, it is a bit more complicated. Just as some
large offices use private branch exchanges (PBXs) so that
individual phones have extension numbers and not unique
telephone numbers, many large companies use a similar
scheme so that the company itself has an IP address and
individual companies have what are essentially extension
numbers, which are used within the company’s internal net-
work.
Here is another way to think about the transition from
8 The Five Eras of Computing

IPV4 to IPV6. IPV4 provides about 4 billion different ad-

dresses, which was originally more than enough to create a
world wide network of computers, but which today is run-
ning out of space. IPV6 provides about 340 trillion, trillion,
trillion (3.4×1038 ) addresses, which is enough today to cre-
ate a world wide network of devices, but which in twenty
years may not be enough. With IPV6 there are enough ad-
dresses so that your phones, games, cars, and cameras can
each have several different addresses.
The transition from a world wide network of computers
to a world wide network of devices is natural and grad-
ual and part of broader transition that is over fifty years
old. Indeed, over the past fifty years, computer hardware,
computer software, and networking have progressed in a
relatively predictable fashion over a trajectory which will
soon create a world in which your cell phone and car are
just as much part of a network as your Gmail account.
If this seems surprising, it is simply because news about
computing is reported as consisting of a series of break-
throughs. In fact, computing can more fruitfully be thought
of as a trajectory in which the slope changes from time to
time as the result of innovation. Innovation is hard to pre-
dict but is quickly recognized and does not change the fun-
damental pattern. This doesn’t mean everything is easy to
predict in computing. For example, predicting which com-
pany will supply the component technology for your cell
phone and car is not so easy. Predicting vendors in this
fashion perhaps may be best considered as the result of a
random walk. News about which vendors are ahead and
which are behind is part of the market clutter which makes
learning about technology confusing. On the other hand,
unless you have invested in one of these companies, you
don’t care all that much as long as one company survives.
We are currently in an era of computing dominated by
web browsers surfing the Internet. Over the next decade
we will transition to an era of computing in which large
numbers of devices will all connect to networks, many of
which will be wireless, and all of which will be able to com-
1.4 The SAD History of Computing 9

municate with each other. Over the past decade we have

emerged from an era of computing characterized by PCs us-
ing office applications, such as word processors and spread
sheets. Prior to that, computing was dominated by termi-
nals connected to mainframes. These four eras span over
half a century. From this perspective, change has been rel-
atively easy to predict. This is the perspective of this book.
In this chapter, we describe these four major eras of
computing that span the past fifty years in more detail.
We will also describe an emerging fifth era of computing
in which data and information will become just as com-
moditized as have computer cycles, storage, bandwidth and
software in the proceeding eras.

1.4 The SAD History of Computing

The river where you set your foot just now is gone —
those waters giving way to this, now this.
Heraclitus
What has been, that will be; what has been done, that
will be done. Nothing is new under the sun. Even the
thing of which we say, “See, this is new!” has already
existed in the ages that preceded us.
Ecclesiastes

In this book, we take a broad perspective on computing

and communication, trying to understand digital comput-
ing and communication from the perspective of decades,
rather than the perspective of a media cycle, in which
there is pressure for each new print edition or each new
broadcast to announce something innovative. To do this,
in Section 1.9 we will divide computing and communication
technologies into five eras: the mainframe era, the personal
computer era, the web era, the device era, and the data
era.
Before we begin though, it is instructive to take an even
longer perspective and to divide computing and communi-
10 The Five Eras of Computing

cating technologies into several epochs, the last of which is

the digital epoch. (Think of epochs as much longer than
eras.)
From this broader perspective, progress in computing
and communications can be viewed along three dimensions:
the symbols used, the algorithms which govern how we ma-
nipulate the symbols in order to compute something, and
the devices used to manipulate the symbols. It might be
helpful to remember the acronym SAD for Symbols, Algo-
rithms and Devices.
A good example of this perspective is provided by the
slide rule. The symbols used by slide rules are the same
symbols that children learn in elementary school today.
What is important to remember is that the positional num-
ber system in which 1402 = 1 x 1000 + 4 x 100 + 0 x 10
+ 2 was not available to the Greeks, and was only fully
developed in India during the 9th century. Because of this,
computation was much more difficult for the Greeks.
Algorithms for multiplying and dividing numbers using
slide rules rely on two formulas involving logarithms that
developed in the 17th century and turn multiplication and
division into addition and subtraction.
A simple slide rule consists of two ruled pieces of wood,
one of which can be moved relative to the other. This is the
device and it is a simple but effective analog computer. The
slide rule as a computing device is so elegant and effective
that it dominated computation for over 300 years.
As another example, today’s digital computers use sym-
bols representing binary numbers, employ algorithms that
add, multiply, and perform other operations on these sym-
bols, and rely on devices that implement these algorithms
using transistors and integrated circuits. Progress is typi-
cally measured simply by how many operations can be per-
formed in one second. In some sense this is like measuring
the beauty of a painting by how many brushstrokes are
used.
From a broad (SAD) perspective, as we will see, much
of today’s computing infrastructure is based upon symbols
1.5 Why Symbols Matter 11

from the 18th century, algorithms from the 19th century,

and devices from the 20th century. Of course the marketing
is from the 21st century, so people today can be proud of
something.

1.5 Why Symbols Matter

All is Number.
Attributed to Pythagoras
Euclid alone
Has looked on Beauty bare.

Edna St. Vincent Millay (1892–1950)

Computation is intimately tied up with the symbols we

use. Symbols and computation are so interwoven that it
is easy to take for granted the power provided by innova-
tive symbols. In this section, we discuss a few examples of
how innovations involving symbols can dramatically sim-
plify computations.
By the time a student today enters high school, she has
written both large numbers (e.g. 299, 792, 458) and frac-
tions (e.g. 3.1415926535) using positional notation; she has
used scientific notation (e.g. 6.38 × 1027 ) for computations;
and she has used symbols to represent numbers (e.g. let x
be a number whose square is 16) and geometric quantities
(e.g. let x be the radius of a circle whose circumference is
10).
What a student doesn’t always appreciate is how rel-
atively recent some of these innovations are: for example,
the transition in Europe from Roman numerals to the posi-
tional Hindu-Arabic system took several centuries and did
not become common until the 16th century; logarithms
were not introduced until the 17th century; and symbols
to represent numeric and geometric quantities were also in-
troduced in the 17th century.
12 The Five Eras of Computing

In classical antiquity, the Babylonians, Greeks, Egyp-

tians, and Romans each used different symbols and number
systems. The Babylonians used a positional number system
with a base of sixty, which we have inherited to this day for
measurements involving navigation, astronomy, and time.
For example, there are sixty seconds in a minute and sixty
minutes in an hour. The sun was observed to require about
360 days to complete a circle, so a circle was divided into
360 degrees, and each degree was divided into 60 minutes
(’), each minute into 60 seconds (”), and each second into
sixty thirds (”’).
Archimedes (c. 287 BC – c. 212 BC) did not have the
use of the positional number system. He wrote a paper
called the Sand Reckoner in which he tried to estimate the
number of grains of sand that would fill the earth. It was
not an easy computation and would have been much simpler
if he had used the positional number system that we take
for granted today.
The Greeks during the time of Archimedes used the let-
ters to represent number as follows: 1, 2, 3, . . . , 9 were rep-
resented by the letters alpha through theta; 10, 20, 30, . . . 90
were represented by the letters iota through koppa (koppa
is not part of the current Greek alphabet); 100, 200, 300,
. . . , 900 were represented by the letters rho through san
(san is not part of the current Greek alphabet). For exam-
ple, the number 222 was written sigma kappa beta.
To represent numbers greater than 999, subscripts and
superscripts were used. For example, adding the Greek
letter iota as a subscript or superscript to the letters al-
pha, beta, . . ., theta, produced the numbers 1000, 2000,
. . . , 9000.
Rules for adding and multiplying using this type of al-
phabetic system were more complicated than the familiar
rules today. Just think for a moment how hard it would be
to estimate the number of grains of sand in the earth using
this number system.
Numbers in the fifth century BC were thought of geo-
metrically. This way thinking can be seen in Euclid’s El-
1.5 Why Symbols Matter 13

5
4

3 3 3

Figure 1.2: The Greek Parthenon is 69.5 meters longs, 30.88

meters wide and 13.72 meters tall, exactly the same pro-
portions that you get if you take three 3x4 rectangles and
lay them end to end as indicated. Note that the length of
the diagonal of a 3x4 rectangle is 5.

ements, which captured the geometry of the 4th century

B.C. Greeks. This was a tremendous achievement and pro-
vided fundamental insights and algorithms for a variety of
different problems, both theoretical and practical.
A good example of how the Greeks viewed numbers as
geometric lengths and ratios is provided by the Parthenon
[115]. The Parthenon is 69.5 meters long, 30.88 meters
wide, and 13.72 meters high. This means that the ratio of
the width to the length is 30.88/69.5 or about 4/9, while
the ratio of the height to the width is 13.72/30.88 or about
4/9. These are the dimensions you would get if you took
three rectangles of length 3 × 4 and placed them side by
side, as in the diagram below. Note that the diagonal of
a 3 × 4 rectangle is of length 5, since by the Pythagorean
Theorem 3 × 3 + 4 × 4 = 5 × 5.
Another significant advance in computing which is easy
to take for granted is the introduction in the seventeenth
century of symbols for unknown quantities, such as the vari-
able x. With these types of symbols, equations such as 2.2
x = 32.8 could be easily represented, as could geometric
objects such as the circles, ellipses, and hyperbolas.
Just as today’s positional number system enables com-
14 The Five Eras of Computing

putations that would be simply impractical with the Greek

or Roman number system, it is instructive to imagine new
types of symbols that might provide a similar advantage
over today’s use of binary symbols to represent positional
numbers.

Here is a simple example. For over thirty years, com-

puter scientists have been building systems for what is
called symbolic computation. These systems manipulate
symbolic as opposed to numeric entities. Using such a sys-
tem, one can multiply two polynomials like x + 2 and x + 3
to compute their product x2 + 5x + 6. Simple versions of
these systems are now found in calculators.

Another example is in logic programming, where sys-

tems work with assertions and rules to draw conclusions
from them.

In Chapter 5, we discuss some of the ways multimedia

digital data, such as images and audio files, are created.
Working with these digitally requires symbols for encoding
images and sounds, as well as devices for taking light waves
and audio waves and producing discrete symbols, such as
the symbols (135, 206, 235) or #82CAFF to represent the
color sky blue in HTML documents.

Today we use symbols from different alphabets to form

expressions which fill files to record a wide variety of things,
including shapes (vector graphics SVG files), colors (JPEG
files), sounds (MP3 files), and video (MPEG and DVD
files).

From this perspective, our ability to compute depends

critically upon the richness of the symbols we use and the
power of the algorithms we use for transforming the sym-
bols.
1.6 Algorithms as Recipes 15

1.6 Algorithms as Recipes

for Manipulating Symbols
Science is what we understand well enough to explain to
a computer. Art is everything else we do.
Donald Knuth

Having fixed a collection of symbols, the next ques-

tion is what do we do with them? What are the rules
and systems for manipulating them? Algorithms may be
thought of as formal procedures for manipulating symbols.
For many problems, such as predicting the trajectory of a
projectile, factoring prime numbers, or predicting tomor-
row’s weather, algorithms affect the speed of computation
just as much, or more, than the hardware of the computing
platform.
It is easy to explain the basic idea of an algorithm using
an example. Recall that x is called a square of a in case x
times x is equal to a. For example, 2 is the square root of
4, 3 is the square root of 9, and 5 is the square root of 25.
Here is a simple algorithm for computing the square
root of a number a.

1. Begin with a guess x for the square root.

2. Replace x by the average of x and a/x.

3. Go to Step 2.

Here are some approximations to the square roots of 2,

5 and 5, 934, 939 using a five line Python program that you
can find in the notes:

sqrt of 2
1.5
1.41666666667
1.41421568627
16 The Five Eras of Computing

1.41421356237
1.41421356237
...

sqrt of 5
2.25
2.23611111111
2.23606797792
2.2360679775
2.2360679775
...

sqrt of 5,934,939
1483735.75
741869.874999
370938.937486
185477.46863
92754.7333983
46409.3593468
23268.6208599
11761.8413875
6133.21703089
3550.44423902
2611.0244305
2442.02762481
2436.18004142
2436.17302342
2436.17302341
2436.17302341
...

The three dots indicate that the last number repeats.

As can be seen from this example, an algorithm consists
of a series of steps each of which can be carried out explicitly
once the previous steps are completed. More or less, one
can think of an algorithm as anything that can be expressed
by a computer program.
1.7 Computing Devices 17

The notes for this section contains another algorithm

called Newton’s Method, which can be used to find the
solutions of a wide class of equations.

1.7 Computing Devices

[Mathematical] Tables have been with us for some 4500

years. For at least the last two millennia they have been
the main calculation aid, and in dynamic form remain
important today. Their importance as a central
component and generator of scientific advance over that
period can be underestimated by sheer familiarity. Like
other apparently simple technological or conceptual
advances (such as writing, numerals, or money) their
influence on history is very deep.
The History of Mathematical Tables: From Sumer to
Spreadsheets, edited by M. Campbell-Kelly, M. Croarken,
R. Flood and E. Robson.

It is easy to take paper, pens and pencils for granted,

but they are important computing devices. Twenty five
hundred years ago, Babylonian computation was done in
clay, Chinese computation was done using bark and bam-
boo, and Egyptian computation was done using papyrus.
For perspective, Homer’s Iliad and Odyssey were not
initially written down, but passed on instead as an oral
tradition. Indeed, they were composed at a time when the
Greek alphabet was still emerging and when papyrus was
scarce.
Papyrus was an early form of paper made from the pa-
pyrus plant, which grew mainly in Egypt and had to be im-
ported by other regions. Egyptian papyri document some
of the earliest algorithms that we know of. For example,
what is sometime called the Rhind papyrus was written
about 1650 BC by the scribe Ahmes. It is about 20 feet
long by 1 foot wide and contains 87 problems, including
problems requiring multiplying numbers and working with
18 The Five Eras of Computing

fractions. For example, Problem 4 shows how to divide 7

loaves of bread between 10 men.
Parchment, which is made from animal skins, was also
used for writing, and may have been invented in part be-
cause of the difficulty obtaining papyrus outside of Egypt.
Parchment can be made from a number of animal skins,
including those of calves, sheep or goats. Although hu-
mans have used animal skins since paleolithic times, the
preparation and use of animal skins for writing is much
more recent. Parchment was invented about third or second
century BCE in Pergamum (modern day Turkey). Parch-
ment became generally available in the Hellenistic world
during the first century AD. Parchment was used exten-
sively through the middle ages, and is still sometimes used
to this day for diplomas or other special documents.
Euclid’s Elements is one of the most important books
containing Greek mathematics. Early versions were proba-
bly written in papyrus, with later versions written in parch-
ment. The earliest version still extant is in parchment and
was written in 888 AD, almost 1200 years after the original.
Paper was invented in China about 105 AD. Paper is
made from fibers extracted from wood pulp, from trees such
as spruce or pine trees. Paper can also be made from fibers
extracted from other sources, such as cotton or rice. In Eu-
rope, paper began to replace parchment during the middle
ages.
Moving from clay and wood to parchment and paper
was one of the first significant advances in computing de-
vices. On the other hand, the Babylonian computations
preserved on clay proved to be more durable, which is why
we know a bit more about how the Babylonians computed
during this time than how some of their contemporaries
did.
Paper is still a wonderful device for computing. It is one
of the most flexible devices and supports not only numeri-
cal computation, but also computation involving algebraic,
geometric and logical symbols. Over the last few hundred
years, it has been augmented by wooden devices, such as
1.8 Case Study: Slide Rule 19

slide rules; mechanical devices, such as adding machines

built from gears; and electronic devices, such as a calcula-
tors and computers built from vacuum tubes, transistors,
and integrated circuits.
There is no reason to expect the development of new
computing devices to slow down. Not only do integrated
circuits continue to improve, but so does the exploration of
novel devices, such as devices that use genes to compute or
that exploit quantum mechanics.

1.8 Case Study: Slide Rule

When I was research head of General Motors and wanted
a problem solved, I’d place a table outside the meeting
room with a sign: LEAVE SLIDE RULES HERE! If I
didn’t do that, I’d find some engineer reaching for his
slide rule.

Charles F. Kettering (1876-1958)

The slide rule is a computing device that enormously

sped up the computation of a sequence of multiplications
and divisions, as well as a variety of other computations,
such as extracting square roots. Slides rules were intro-
duced in the 17th century and were an important comput-
ing device for over three hundred years, a reign significantly
longer than the modern digital computer.
Prior to slide rules, the device most often used for math-
ematical computations was probably a mathematical table.
Mathematical tables are one of the earliest computing de-
vices. We have examples of tables used by Babylonians
dating from about 2000 BC.
Tables are still used in mathematics to this day, and so
are in the running for one of the computing devices with the
longest staying power. Tables are easy to use. For example,
using the table of sines below, we can read off the sine of 80
degrees as 0.98481 by reading the table from left to right.
We can also read the table from right to left to see that the
20 The Five Eras of Computing

c=axb C=A+B

Step 1. Compute A = f(a), B= f(B)

Step 2. Compute C = A + B
Step 3. Compute c = g(C)

Figure 1.3: The first idea underlying the slide rule is that
there are certain functions (f and g in the diagram) that
transform multiplication and division into addition and
subtraction. John Napier discovered such a pair of func-
tions in the early 17th century.

number whose sine is −0.34202 is 200 degrees. Reading the

table from right to left computes a function which undoes
the sine (the function is called the arcsine). To compute
the sine of 50 degrees, we know from the table that it is
between 0.64279 and 0.86603 (the sines of 40 degrees and
60 degrees). If we take the midpoint of these two numbers
we get 0.75441, which is close but not exactly the sine of
50 degrees which is 0.76604. Over the years, a number of
formulas have been developed to interpolate between two
values in a table to get more accurate answers.
There are two basic ideas underlying the slide rule:
The first idea is that certain functions can reduce multi-
plication and division to addition and subtraction. In 1614,
John Napier (1550-1617) introduced a very nice function (a
variant of today’s logarithm function) with the property of
reducing multiplications to additions. See Figure 1.3.
The second idea is due to Edmund Gunter (1581-1626).
1.8 Case Study: Slide Rule 21

degrees sine
0 0.00000
20 0.34202
40 0.64279
60 0.86603
80 0.98481
100 0.98481
120 0.86603
140 0.64279
160 0.34202
180 0.00000
200 -0.34202
220 -0.64279
240 -0.86603
260 -0.98481
280 -0.98481
300 -0.86603
320 -0.64279
340 -0.34202
360 0.00000

Table 1.1: This is a simple mathematical table. If you

read from left to right, you can compute the sine of various
numbers. If you read from right to left, you can compute
another function, called arcsine, that is the inverse of the
sine function.
22 The Five Eras of Computing

1 2 3 4 5 6 7 9

Figure 1.4: Edmund Gunter positioned numbers along a

piece of wood according to their logarithms. The Gunter
scale was used by seamen to simplify computations for nav-
igation.

1 2 3 4 5 6 7 9

1 2 3 4 5 6 7 9

Figure 1.5: About 1625, William Oughtred put together

two Gunter scales, allowing one to slide next to the other
to create the slide rule.

His idea was to create a scale on wood in which numbers

were marked according to their logarithms. See Figure 1.4
About 1625, William Oughtred (1575-1660) put together
two Gunter scales so that one could slide next to the other,
creating the slide rule. Slide rules were one of the staples of
computation until the introduction of handheld calculators
in the 1970s.
To end the section, let us try to characterize the slide
rule from the symbols-algorithms-devices (SAD) perspec-
tive:
First, let us think of the slide rule as a device. From this
perspective, it was a true revolution. Prior to slide rules,
computations were done using tables, for example, tables
of sines, cosines, square roots, and logarithms. The slide
rule allowed one device (mathematical tables) and several
manual look-ups to be replaced by another device (the slide
rule) and two operations (moving one wooden rule with
respect to the other) and moving a reference line along the
1.9 From Mainframes to Devices 23

two slides. In other words, this elegant device replaced large

mathematical tables and error prone manual look-ups. In
fact, multiple tables could be encoded on the same device,
so that a single slide rule could replace tables of square
roots, cube roots, and logarithms.
Second, consider the algorithms involved. From this
perspective, the slide rule is less innovative. Indeed, the
algorithms used for a slide rule are essentially the same
algorithms used previously with mathematical tables. The
first algorithm is a simple look-up (reading a table from left
to right) or reverse look up (reading a table from right to
left). The second algorithm is using a function like the log-
arithm to replace multiplications and divisions by additions
and subtractions.
Third, consider the symbols involved. Again, from this
perspective, the symbols used in mathematical tables and
the symbols used in slide rules are essentially the same:
positional numbers (like 1.414213562).
Although the computing devices we describe next don’t
have the staying power of mathematical tables (4000+ years)
or slide rules (300+ years), they do have the advantage
that they can also be used for playing music and writing
emails to your family and friends. On the other hand, nei-
ther mathematical tables nor slide rules require batteries or
electricity and both function perfectly well a decade after
production.

1.9 From Mainframes to Devices

The longer you look back, the further you can look
forward.
Winston Churchill
24 The Five Eras of Computing

It is difficult to tell a short-sighted man how to get

somewhere. Because you cannot say to him: “Look at
that church tower ten miles away and go in that
direction.”
Ludwig Wittgenstein

One of the themes of this book is that digital computing

for the past fifty years or so has been shaped in large part by
the process of commoditization. To understand commodi-
tization better, it is useful to divide the past fifty years into
four eras, consisting of overlapping 20 year periods.

1. The Mainframe Era (1965 - 1985). In 1965, IBM

shipped the System 360, its first computer based on
integrated circuits. By 1968, it had installed over
14,000 System 360 systems, at an average price of
over $1,000,000 per system, generating over $14 bil-
lion dollars of revenue for IBM. During the Main-
frame Era, computing cycles were limited. Depart-
ments were built around the mainframes to operate
them, ration their cycles, and provide services to the
users.

2. The Personal Computer (PC) Era (1980-2000). In

1977, Apple, Commodore and Tandy began selling
personal computers. In 1981, IBM entered the mar-
ket, effectively setting the standard, with a business
model which encouraged third party manufacturers.
For example, Compaq shipped its first IBM clone in
1983 and set a record by selling $111 million of PCs,
one of the largest first year product sales in the his-
tory of American business. By the end of this era,
over 50 percent of the workers in some metropoli-
tan areas had PCs. PCs had also became building
blocks for specialized computing needs: for example,
rather than build specialized devices for supercom-
puting, supercomputers were being built by putting
together hundreds or thousands of PCs into what are
1.9 From Mainframes to Devices 25

called clusters. During the PC Era, computer hard-

ware became a commodity.

3. The Web Era (1995-2015). In 1993, Mosaic, the first

graphics based web browser, was released. It was de-
veloped at the National Center for Supercomputing
Applications at the University of Illinois in Urbana
Champaign. In 1995, NSF turned over the network-
ing backbone for the Internet to commercial vendors
and, at the same time, introduced a research network
called the very High Speed Backbone Network Ser-
vice or vBNS, that became the foundation for the
next generation of the Internet. By 1996, the num-
ber of hosts on the Internet exceeded 10 million. By
2000, the number exceeded 75 million hosts. With
the Internet, it became much easier to develop and
deploy software applications that essentially used the
Internet as the operating system. Companies began
to give away software. For example, Hotmail was a
free email program and had over 30 million users two
and half years after its launch in 1996. During the
Web Era, software became a commodity.

4. The Device Era (2005-2025). There is no generally

accepted name for the post-web era. In this era,
PCs connected to the web will be supplemented by a
wide variety of networked devices, many of which are
wireless. These devices include mobile phones, per-
sonal digital assistants (PDAs), devices for listening
to music, cars emitting diagnostic information, cam-
eras, home stereos, as well as a variety of other devices
not yet prototyped. At the beginning of the device
era, wireless Internet is available at coffee shops and
bookstores, and some cities are providing free wire-
less Internet. During the Device Era, networks will
become a commodity.

Dividing the digital age into four eras simply provides

convenient markers for certain inflection points. For ex-
26 The Five Eras of Computing

ample, although the web era is defined as the twenty year

period from 1995-2015, the roots of the web era date back
at least to the sixties. For example, the first paper on
packet switching, which is one of the main ideas for the
Internet’s network architecture, was published by Leonard
Kleinrock from MIT in 1961. In 1967, Larry Roberts of
ARPA organized the first meeting to design what would
become ARPANET, the predecessor of the Internet. In
1969, ARPANET began operations with nodes at UCLA,
the Stanford Research Institute (SRI), the University of
California at Santa Barbara (UCSB) and the University of
Utah.
In other words, the divisions between the different eras
should be thought of as simply convenient markers denoting
a time in which the prior era’s technology is widely diffused,
and the next era is well on its way. For example, by 1996,
near the beginning of the web era, there were over 10 million
hosts on the Internet. By this time, your father and your
grandfather were likely to have heard of the Internet.

1.10 The First Era: Mainframes

From today’s viewpoint, life with the first computers seems
to belong to a simpler, more romantic time: there was ba-
sically one type of computer — the mainframes — and one
dominant company — IBM. Of course, if you look deeper,
there were specialized computers for specialized purposes,
and several companies trying to gain the dominant market
share.
More importantly, though, the computing was central-
ized, and the control of the system was firmly in the hands
of the professional computing staff. Individuals accessed
information through terminals that were networked to the
mainframe. To get new information from the system or old
information presented in a new way, one submitted a writ-
ten request which eventually either led to what was wanted
or to a further request. On the positive side, the mainframe
1.11 Case Study: Punch Cards 27

with its centralized structure meant that the information

was secure and that operations were managed by profes-
sionals. One also knew who to blame when things went
wrong.
During this period, mainframe computers primarily man-
aged corporate level information, such as employee records,
accounting data, and product data.
In a sense that will become clearer as we consider the
second and third eras, the first era can be characterized as
being hardware-limited. The challenge was to provide the
appropriate hardware to the customer. Companies such
as IBM, which provided this hardware and the associated
services, grew to be quite large.

1.11 Case Study: Punch Cards

Old technologies are a bit like photographs — you never re-
ally throw them away, you just spend less time with them.
In the second and third eras, the era of PCs and the Web,
many companies still use mainframes for important busi-
ness processes, such as the preparation of payrolls.
Not only are mainframes still critical in producing pay-
rolls but so are punch cards. In the 1880’s, the punch
card was invented by Herman Hollerith to automate the US
Census. Holes in a punch card are translated by a punch
card reader into mechanical or electrical signals that can
be counted or further processed.
One of the most common uses of punch cards was to
compute payroll. For example, the garment industry used
punch cards as follows. A punch card was created each
time an individual working on a garment performed a par-
ticular operation in the manufacture of the garment. If
a worker performed 100 operations per day for five days,
this generated 500 punch cards. One hundred employees
would generate 5000 punch cards per week, which would
be tabulated to produce the week’s payroll.
In the 1970’s, disk drives began replacing the punch
28 The Five Eras of Computing

cards as the main storage mechanism for data. Today, it

is very difficult even to find a punch card reader. On the
other hand, systems using them as the basic form of data
entry still persist. In 2002, the New York Times ran a story
describing how the 23,000 employees of the Los Angeles
County Department of Health Services still use punch cards
for their time card and payroll system [83]. Punch cards
are still used because the system still works and moving
all the hospitals and clinics to a new system would be too
expensive.
In 1967, an estimated 260 billion punch cards were used
in the US or 1,300 for each American. Today, there are
only three companies in the US that manufacture punch
cards. On the other hand, punch cards are still used for
many voting machines, as became apparent in the 2000
Presidential election controversy.

1.12 The Second Era: PCs

The emergence of microprocessor-based computers ushered
in a new era because for the first time hardware became a
commodity. During this era, personal computers covered
more and more desk tops. Companies that provided useful
desktop applications such as word processors, spreadsheets,
and databases grew. Hardware was distributed and com-
puting was based on the client-server model in which larger
systems (servers) provided data and information to smaller
systems, such as personal computers (clients) networked to
them.
Individuals could suddenly buy, install, and use desktop
applications without the lengthy involvement of a central-
ized computing department. On the other hand, individ-
uals were suddenly responsible for managing and servicing
their desktop systems and protecting their desktop data.
The information on the desktop was, by and large, de-
partment level information generated by individuals: re-
ports, presentations, budgets, and other work related data.
1.12 Personal Computer Era 29

The VisiCalc spreadsheet was one of the first new types

applications to be designed for the PC. VisiCalc was de-
signed and developed by Dan Bricklin and Bob Frankston.
Bricklin and Frankston formed Software Arts, Inc. on Jan-
uary 2, 1979, and worked out of their attics. There was
very little infrastructure for developing applications on the
PC in those days. The first version of VisiCalc was written
for the Apple II in assembler, a low level language that re-
quired quite a bit of code for such things as reading input
from the keyboard, saving information to a file, or reading
information from a file. With higher level languages these
days, these operations can all be done with a single line of
code.
If you grew up using spreadsheets, it may be a bit diffi-
cult to imagine how business was conducted before they
were introduced. The first business application of Visi-
Calc took place while Dan Bricklin was getting an MBA
at the Harvard Business School in 1979. MBAs at Harvard
learned about business through case studies. Bricklin used
an early version of VisiCalc to analyze the Pepsi Challenge
marketing campaign for the Pepsi Corporation case study.
While the other students in the class used a Texas Instru-
ment Business Analyst calculator, Bricklin used VisiCalc,
allowing him to do five year forecasts and vary many of the
assumptions. When the professor teaching the class asked
Bricklin how he had done all the projections, Bricklin an-
swered in a way that didn’t disclose that he had used an
entirely new type of product, since he wanted to keep the
VisiCalc secret until it was closer to being released.
A few years later, spreadsheets would transform many
business processes. For example, Kohlberg Kravis Roberts
& Company, commonly referred to as KKR, is a New York
City-based company that popularized the leveraged buy-
out. One of their secret weapons was using spreadsheets to
analyze quickly the cash generated under various scenar-
ios when the acquired company was split apart in different
ways. This was very important, since KKR used the ac-
quired company’s own cash for the buyout, and financed
30 The Five Eras of Computing

the rest by issuing high yield (also known as junk) bonds.

In some sense, the PC Era was limited by the availabil-
ity of the appropriate desktop software and the challenge
was to provide this software to the customer. Companies
such as Microsoft that met this challenge became the icons
for this era.

1.13 The Third Era: The Web

A typical software application during the PC Era was Mi-
crosoft Word, a word processor. A typical software appli-
cation in the Web Era is Google’s GMail, an email pro-
gram. Software from the PC Era resides on your desktop
computer or your laptop. Software from the Web Era is
often on another machine and you access it through a web
browser over a network.
The significance of the world wide web is not only that
one can click to bring up an interesting picture or that one
can buy a book over the web, but rather that information
and services from millions of servers on the network are just
as easily available as if they were your own desktop PC.
The icons of this age are not the hardware vendors —
IBM sold its PC business in 2004 — nor the software ven-
dors — much of the software infrastructure, such as the
Apache web server, is open source and not sold by a soft-
ware vendor but instead developed by a distributed team
of volunteers — but rather providers of web-based services,
such as Amazon, eBay, and Google.
Think of it this way: someone buying a book on Ama-
zon is more aware of how slow her network connection is
than what the model of her PC is or which software ven-
dor developed her web browser. Over a two year period,
the book buyer will easily pay more for her DSL-based net-
work connection (2 years x 12 months per year x 50 dollars
per month or $1200) than for her PC clone ($800) or her
open source Mozilla Fire Fox web browser ($0). During
the web era, hardware and software are commodities, but
1.14 Case Study: SMTP 31

network access and web based business services are still a

scarce resource.
During the web era not only is network access a scarce
commodity but so is information. This is a bit counter-
intuitive, since suddenly not only is everything on your own
computer accessible, but, via the web, so is much of the in-
formation on the computers of colleagues and strangers.
The problem, though, is that this information is rarely or-
ganized well enough so that you can readily use it.
Think for a moment about how information is organized
on your own desktop or laptop computer. For most of us,
organizing information on our computer is put on a to-
do list almost as frequently, and no more effectively, than
losing ten extra pounds of weight. Suddenly, not only must
your computer be organized, but so must the computers of
all your colleagues, as well as those of strangers you have
never met.
In other words, although the amount of digital data
is growing, the amount of useful information is not keep-
ing up. The technology to screen, sort, and extract useful
information from large amounts of digital data is not yet
ready for everyday use. This is the challenge of the next
era. Companies that succeed with these challenges, such as
Google, are candidates to be the icons of the next era.

1.14 Case Study: SMTP

This section contains a case study of the Internet protocol
that powers email. An Internet protocol is roughly speaking
a “language” that allows two computers on the Internet
to communicate. There are a large number of different
Internet protocols, including those that support email, web
browsing, and setting up telephone calls over the Internet.
Email is commonly considered as the killer application or
killer app that sparked the adoption of the Internet among
consumers.
Many users access email through an email application
32 The Five Eras of Computing

such as Microsoft’s Outlook or Outlook Express. Outlook is

a proprietary client that not only interfaces to proprietary
Microsoft products and services but also interfaces to open
protocols and services, such as the Simple Mail Transfer
Protocol or SMTP. Although most people have never heard
of it, SMTP is one of the main reasons that email became
a killer app.
SMTP is a computer-to-computer (C2C) protocol that
allows two computers on the Internet to exchange email
messages [126]. Here is an example from the original 1982
description of SMTP [126]:
S: MAIL FROM:
R: 250 OK

S: RCPT TO:
R: 250 OK

S: RCPT TO:
R: 550 No such user here

S: RCPT TO:
R: 250 OK

S: DATA
R: 354 Start mail input; end with .
S: Blah blah blah...
S: ...etc. etc. etc.
S: .
R: 250 OK
EMail is exchanged when an email client connects to a
SMTP server. The SMTP server is a computer that pro-
vides an Internet or web service. In particular it listens for
requests from SMTP clients. Clients include proprietary
Microsoft products such Outlook, as well as browser-based
email, such as those provided by Google, Yahoo or Hotmail.
In the computer-to-computer conversation above, Smith
(the sender S) at the computer alpha.arpa tries to send
1.14 Case Study: SMTP 33

email to Jones (the receiver R) at beta.arpa and Green

(the receiver R) at beta.arpa. He succeeds with Jones but
not with Green. The email is sent as an ASCII text stream.
The computer-to-computer conversation takes place be-
tween two different computer programs residing on two dif-
ferent nodes on the Internet without human intervention.
A special case is when both programs reside on the same
node. The conversation between the SMTP-Sender and the
SMTP-Receiver is pretty simple and consists of three steps.
In the first step, the SMTP-Sender sends a mail com-
mand:

MAIL FROM:

This command tells the SMTP-Receiver that a new

message is being sent. The reverse path is used so that
the SMTP-Receiver can return a 250 OK reply after suc-
cessfully processing the command. In the second step, the
SMTP-Sender sends a RCPT command:

RCPT FROM:

This command gives the address of the one recipient of

the message. If the SMTP-Receiver accepts the recipient,
it returns a “250 OK”; if not, it returns a “550 No such user
here”. This step is repeated once for each user the message
is sent to.
In the third step, the SMTP-Sender sends a DATA com-
mand containing the message itself:

DATA

If the SMTP-Receiver accepts this commands it returns

a “354 Intermediate Reply”. Once the SMTP-Sender re-
ceives the 354 command, it begins sending data as a simple
ASCII stream. The end of the ASCII text stream is indi-
cated by sending a single line consisting of a single period
(“.”) followed by a carriage return and line feed. Once the
SMTP-Receiver receives the period, it sends a final 250 OK
command to finish the computer-to-computer session.
34 The Five Eras of Computing

Notice that the FROM, TO, DATE, SUBJECT, CC,

BCC, and other fields in a standard email message are all
sent as part of the data and have no special meaning to
the SMTP-Sender or SMTP-Receiver. The mail client pro-
gram, such as Outlook or a Yahoo web mail, extracts the
mail addresses from these fields and passes them to the
SMTP-Sender.
Over time, mail was used to send a variety of attach-
ments in a variety of formats, from Microsoft Word doc-
uments to jpg images. To handle this, a standard was
developed called the Multimedia Internet Mail Extensions
(MIME) Encoding. MIME is a way for binary data such
as Microsoft Word documents or graphic images to be en-
coded as ASCII text. Once encoded in this way, it can be
sent using standard SMTP-Senders and SMTP-Receivers.
The same MIME encoding is used today by both SMTP
and HTTP.

1.15 The Fourth Era:

Clouds of Devices
In the fourth era, hardware cycles and software applica-
tions have already become commodities. This era is about
the transition from browser-based applications to devices;
from copper-based networks to fiber-based and wireless net-
works; and from an application-based software model to a
service-based software model.
The icon of the third era is a PC running a web browser.
The problem is that 1990’s style PCs were complicated to
operate, had to be tethered to networks, and came in ba-
sically one style and color. The icon of the fourth era is a
wireless pocket device supporting email and instant messag-
ing. The device doesn’t need to be booted, simply turned
on; it doesn’t need a network cable, simply a network; and
it doesn’t need an operations manual, simply an operator.
In the fourth era, rather than use a single PC running
desk top applications in your office, you are more likely to
1.15 Clouds of Devices 35

carry several small independent devices, such as email de-

vices, cell phones, music and video players, and games, all
connected on wireless IP networks and providing various
services. In the same way that a telephone today provides
a simple service (you dial a phone number and talk to some-
one), these devices will also provide equally simple services
(you enter an email address, a short note, and push send).
By the end of the fourth era computing devices with em-
bedded wireless networking and general positioning systems
(GPS) capabilities will be smaller than postage stamps and
cost about the same. They will be included in automobiles
to provide early warnings of engine breakdown, attached
to bridges to aid in maintenance, and used to keep track of
your young children.
In the first two eras, the computer was firmly at the
center of the model. Attached to the computer were vari-
ous peripheral devices, such as terminals, printers and disk
drives. In the third era this began to change, with the net-
work moving towards the center. By the fourth era, this
transition should be complete, with computer routers and
switches firmly at the center of the model, and with CPUs,
disk drives, and wireless devices simply peripherals. Some-
times this model is called the “hollowed out computer.”
There is no agreed upon name for the fourth era. The
fourth era is full of devices providing clouds of services over
an ever present and ubiquitous IP network. For simplicity,
in this book, we will refer to it simply as the Device Era.
You should expect this name to seem quaint and old fash-
ioned by the time you are reading this book, but I would
be surprised if the essence of the era had changed in a sig-
nificant way.
What is scarce in the fourth era is a way to manage the
data and the information produced by the myriad devices
and their associated services. The upcoming fifth era is
based upon the emerging ability to extract useful informa-
tion from this data and to structure it in a way that leads
to useful decisions.
36 The Five Eras of Computing

1.16 Case Study: Routers

This case study is about a piece of hardware called a router
that allows two or more computer networks to connect to-
gether. The development of the router was one of the crit-
ical events that enabled the creation of the Internet.
Viewing a web page on the Internet requires that your
computer send messages back and forth with another com-
puter, which is called the web server. In an earlier section,
we learned that just as every telephone in the world has a
unique phone number, every computer on the Internet has
a unique address, called its Internet or IP address.
Very roughly, the Internet, from the network perspec-
tive, is a collection of local area networks that are con-
nected together with routers and that communicate with
a common set of protocols. These protocols divide data
into chunks called packets and attach the IP address of the
source and the IP address of the destination to each packet.

As more and more local area networks connected to

the Internet, more and more routers were required. The
company CISCO was an early supplier of routers. It was
founded in 1984. It sold approximately 5,000 routers in
1990 and over 900,000 in 1997. By about 2002, routers
could process millions of packets per second, and by 2006,
they could process billions of packets per second. Routers
contain tables that may have over a hundred thousand en-
tries describing how to route an incoming packet.
Here is how they work in a bit more detail:
Suppose that Computer A in local area network L1
wants to send a message to Computer B on local area net-
work L2. Suppose also that the two local area networks are
connected together with a router R having three ports. In
this context, a port is a number used to distinguish two or
more physical connections of a network to a router. Port
P1 is connected to local area network L1, Port P2 is con-
nected to local area network L2, and Port P3 is connected
to the local area network of an Internet Service Provider.
1.16 Case Study: Routers 37

Here is what router R does every time it processes a packet.

1. Computer A first breaks the message into units which

are each 1500 Bytes long. For simplicity, assume that
the message is short and consists of a single 1500 Byte
data packet.

2. Given the name of Computer B, Computer A looks up

the corresponding IP number for Computer B using
a network service called the Domain Name Service or
DNS.

3. Computer A then assembles a packet [IP(A), IP(B),

data], where IP(A) is the source Internet address of
computer A, say a.b.c.d, IP(B) is the destination In-
ternet address of Computer B, say e.f.g.h, and data
is the data packet for the message.

4. Computer A looks at the destination address e.f.g.h

and determines that it is is not a packet on its own
local area network. If it were, the last part of the
address would be of the form b.c.d. It therefore sends
it to a specific computer called the default gateway,
which is specified in the network configuration soft-
ware of Computer A. Call the default gateway Router
R.

5. To send the packet from A to the Router R, the Com-

puter A wraps the IP packet in a frame of the type
required by the local area network. For example, if
the local area network uses Ethernet, then [MAC(A),
MAC(R), IP(A), IP(B), data, CRC]. Here MAC(A)
and MAC(R) are the MAC address of the network
interfaces for Computer A and Router R on the local
area network L1.

6. When Router R gets the packet, it removes the Eth-

ernet Frame to get [IP(A), IP(B), Data]. Router R
looks at its router table and finds that IP addresses
like B are sent to Port P2 on Router R.
38 The Five Eras of Computing

7. To send the packet from Router R to Computer B on

local area network L2, Router R places [IP(A), IP(B),
Data] into a frame for local area network L2.

8. Computer B on local area network L2 receives the

frame, extracts the IP Packet, and then extracts the
data.

1.17 The First Half Century

of Computing

Study the past if you would define the future.

Confucius

In this section, we broaden our point of view a bit and

consider computing platforms, which include not only the
computer itself, but the broader infrastructure required for
computing, including operating systems, applications, stor-
age devices, networks, network services, displays, peripher-
als, and various other devices that we use when computing.

Viewing computing platforms from the perspective of

fifty years or so is difficult for several reasons:

• The different components of a computer and of a

computing platform become commoditized at differ-
ent rates. This is the subject of Chapter 2.

• We are surrounded by marketing clutter and the amount

and type of clutter varies from component to compo-
nent in the computing platform. This is the subject
of Chapter 3.

• New technology gets adopted over a period of time.

Different factors affect the rate at which different com-
ponents of a computing platform are adopted. This
is the subject of Chapter 4.
1.17 The First Half Century of Computing 39

In this section, we briefly consider each of the difficulties

in turn from the perspective of about fifty years.
Computers vs. Computing Platforms. As we just
mentioned, computers are the most visible component of
the computing infrastructure, but not the only component.
What we do with computers depends upon the software
applications that run on them, the operating systems they
employ, the displays and peripherals they use, the networks
that connect them, the data that flows through them, and
the human-computer interfaces by which we interact with
them. One way to to understand the differences between
different computing platforms is to answer the following
questions:

1. What is the hardware?

2. What is the software?

3. What is the network?

4. What is the user interface?

5. Where is the data?

Table 1.2 summarizes how each computing era has an-

swered these questions differently. It is important to keep
two things in mind:

1. First, there is no sharp division between one era and

the next. Indeed, it is usually several years into the
new era before there is a broad understanding of the
nature of the technology of the new era, and a corre-
sponding disappointment in all the utterances by the
pundits about the transition.

2. Second, the computing platforms from prior eras never

really fade away; instead, they continue to hang around,
continue to be used, and continue to evolve. It is sim-
ply human nature to focus on what is new.
40 The Five Eras of Computing

Scarce and plentiful resources. A simple means of

distinguishing the different computing eras is to ask what
component of the computing platform is the bottleneck,
and hence is rationed by high prices, and what component
is becoming a commodity, characterized by falling prices
and rapidly increasing capacity. This is the subject of the
second chapter. Table 1.3 provides a high level summary.
Technology adoption. Despite their name, computers
do much more than compute. In fact, relatively few people
use them for computing. Today a consumer at home is more
likely to use a computer to send email, to buy books, to play
games, or to listen to music than to perform a computation.
Similarly, a business today is more likely to use a computer
to do their accounting, to pay their bills, to manage their
inventory, to keep the loyalty of their customers, or to create
a marketing brochure.
Each time a new application or function appears, it
usually takes much longer than is initially predicted to be
adopted. In Chapter 4, we examine some issues that arise
and limit the spread of new technology. The term technol-
ogy adoption life cycle is often used to describe the process
that limits the spread of new technology. The good news
is that you can pay consultants to tell you that the adop-
tion of new technology can be challenging. If you pay them
enough, they will tell you that the color of your marketing
brochure is wrong and that the tag line you are using in the
brochure is not catchy enough. Unfortunately, sometimes
the problems are more fundamental.
Question First Era Second Era Third Era Fourth Era
When? 1965-1985 1980-2000 1995-2015 2005-2025
What is the mainframes servers & personal devices
hardware? PCs computers
What is the back office PC applica- web services on
software? applications, tions, such applications, devices, such
such as as word pro- such as as listening
payroll cessors and Amazon and to music or
spreadsheets Facebook taking pic-
tures with
cell phones
What is the terminals on local area wide area wireless net-
network? serial lines networks networks works
What is the terminals PCs with clicking pushing but-
user interface? windows, buttons on tons on de-
menus & browsers vices; voice
mice
1.17 The First Half Century of Computing

Where is the in the main- on servers on desktops in your

data? frames pocket
41

Table 1.2: The first four eras of computing.

 42

Mainframe PC Era Web Era Device

Era Era
What is the computer application network data
bottleneck? cycles software bandwidth
What is be- computer application network
coming com- cycles software bandwidth
moditized?

Table 1.3: Viewing the four eras of computing by what is the

bottleneck and what is a commodity.
The Five Eras of Computing
1.18 The Commoditization of Data 43

1.18 The Fifth Era:

The Commoditization of Data
The majority of books about technology that predict the
future age very quickly and become irrelevant within a few
years. With this in mind, and with the fifth of era of com-
puting just beginning to emerge, it will be difficult to say
very much about it. On the other hand, extrapolating the
perspective of the section above, we can provide a rough
characterization of some aspects of it:

• Just as the second era commoditized cycles and cylin-

ders, the third era application software, and the fourth
era bandwidth, the fifth will commoditize data.

• Once data is commoditized, new types of discovery in

science and new types of decision support applications
in business will become common.

• The scarce resource in the fifth era of computing will

be those individuals with knowledge of how to lever-
age the technology — everything else will be com-
moditized or outsourced.

The Fifth Era of Computing will be the subject of Chap-

ter 5. Chapter 5 will discuss the commoditization of data
and the emergence of a computing platform that promises
to change data-driven decision-making in the same way
that during the Fourth Era, the Apple iPod changed how
teenagers listened to music.