CSE584: Software Engineering

Lecture 1: Introduction & Overview

David Notkin
Computer Science & Engineering
University of Washington
http://www.cs.washington.edu/education/courses/584/
 Outline
• Intent and overview of course
• Software engineering overview
– Stuff you already know, but it’s important to lay it out
so we are working from the same page
• Notkin’s top 10 “insights”
– My goal is to lay out my prejudices and views, to
increase your understanding of the intent of the course
• Overview of course work and
administrivia
 Introductions
• Very useful for me (and you)
– What do you do?
– What do you want from the class?
– What are the most serious
software engineering problems
you face?
• But time consuming, so we’ll do
it electronically
– Through the email list
([email protected])
– Distributed to the entire class
 But I do want some basics
• What companies do you work for?
• What is your general responsibility?
– Development, testing, maintenance, other?
• Take a couple of minutes at each site to
gather these data
– I’ll handle the UW site
– The person whose last name comes first
alphabetically handles the other sites
• Announce when you’re ready
 Interaction
• I like to have interaction with students
during class, especially 584
– You have tons of key insights in your head

 
– It’s boring just listening to me
• Especially in the                                              
evening & during a long class
• Try just interrupting me; if that doesn’t
work, we’ll try something else
• Remind me to repeat questions, since it’s
often hard to hear them at other sites
 Your undergraduate
experience?
• How many of you took This is my guess
an undergraduate
software engineering about your answers
course?
• Did any of you think it
was good?
• What, specifically, was
particularly good or
bad about it?

For non-UW grads,

that is!
 Intent of course
• Most of you have jobs engineering software
– I don’t (and I never really have)
• So, what can I teach you?
– Convey the state-of-the-art
• Especially in areas in which you don’t usually work
– Better understand best and worst practices
– Consider differences in software engineering of
different kinds of software
• You provide the context and experience
• Meeting and talking to each other is key
Lots of differences among you
• You have a lot in common
– Undergrad degree in CS or related field
– Significant experience in the field
– You’re really smart
• You also have a lot of differences
– Development vs. testing
– Desktop vs. real-time
– Different company cultures
– …and much, much more
• This in part will be why some material in the course
will resonate with you, while other material won’t
 My metric for success
I will consider this a successful course if,
over the course of the next year or so, you
approach some specific problem you face
differently because of the course

• Maybe from readings

• Maybe from discussions with other students
• Maybe from assignments
• Maybe even from lecture
 Another key intent

• There is general agreement that

– Research in software engineering doesn’t have
enough influence on industrial practice
– And much in industry could improve
• Why is this true?
– What can academia do to improve the situation?
– What can industry do to improve the situation?
• By the way, I believe that this perception is
not entirely accurate
– But it’s still a crucial issue for lots of reasons
 Possible impediments

• Lack of communication
– Industry doesn’t listen to academia
– Academia doesn’t understand
industrial problems
• Academic tools often support
programming languages not
commonly used in industry
Other possible impediments?

• In groups of 3 or 4, list some

other possible impediments
• In 3 minutes, we’ll gather a set of
suggestions

GO! STOP!
 Tichy’s main impediment
• Walter Tichy has claimed that the major impediment is the
lack of “experiments” in CS research
– “Should Computer Scientists Experiment More? 16 Excuses
to Avoid Experimentation.” IEEE Computer 31(5), May 1998.
– http://wwwipd.ira.uka.de/~tichy/
• I have lots of reactions, including
– I don’t think industry, as a rule, finds this to be a (the) major
impediment
– We do experimentation, in a different style
– Evaluation is difficult in software engineering, so we must be
creative
– This is an example of science envy
Software is increasing critical to
society
40000
35000
30000 CSE143 program
25000 4-speed xmission
20000 ATM machine
15000
B-2
10000
5000
NT5.0
0
KLOC

and it’s getting bigger and more complex

 Absolute sizes
140
ATM (my 3 year-
old)
120
B-2 (Tao on my
100 shoulders)
80 NT5.0 (Statue of
Liberty from base)
60
40
• 50 lines per page
20 • Double sided
• 500 pages/ream
0 (2 inches)
Feet High
 How I spend my time
• The Great Pyramid of Giza is 481’
• The Kingdome was 250’
• The Colossus of Rhodes is 110’
• The Eiffel Tower is 1033’
• The Graduate Reading Room in Suzzallo is 65’
• A 747 is 63’ to the top of the tail
• The Brooklyn Bridge is 135’ above the water
• Titanic’s height from keel to bridge is 104’
• The EE1 building is about 90’
 Delivered source lines per
person
• Common estimates are that a person can
deliver about 1000 source lines per year
– Including documentation, scaffolding, etc.
– Independent of the programming language
– Yeah, you do better 
• Obviously, most complex systems require
many people to build
• Even an order of magnitude increase doesn’t
eliminate the need for coordination
 Inherent & accidental
complexity
• Brooks distinguishes these kinds of
software complexity
– We cannot hope to reduce the inherent complexity
– We can hope to reduce the accidental complexity
• Some (much?) of the inherent complexity
comes from the incredible breadth of
software we build
• That said, it’s not always easy to distinguish
between these kinds of complexity
 “The Software Crisis”
• We’ve been in the midst of a “software crisis”
ever since the 1968 NATO meeting
– crisis — (1) an unstable situation of extreme danger or
difficulty; (2) a crucial stage or turning point in the
course of something [WordNet]
– I was 13, and many of you weren’t born yet
• We cannot produce or maintain high-quality
software at reasonable price and on schedule
– Gibb’s Scientific American article [in your course pack]
– “Software systems are like cathedrals; first we build
them and then we pray” —S. Redwine
Some classic “crisis” issues
• Relative cost of hardware/software
– Where’s Moore’s Law for software?
• Low productivity
• “Wrong” products
• Poor quality
– Importance depends on the domain
• Constant maintenance
– “If it doesn’t change, it becomes useless”
• Technology transfer is (too) slow
 Notkin’s view—“mostly
hogwash”
• Given the context, we do pretty well
– We surely can, should and must improve
• Some so-called software “failures” are not
– They are often primarily management errors (Ariane,
Denver airport, U.S. air traffic control, etc.)
• Interesting recent article in the Wall Street Journal on
Australia’s and New Zealand’s success in air traffic control
– Read comp.risks
• In some areas, we may indeed have a looming crisis
– Safety-critical real-time embedded systems
– Y2K wasn’t
 Software engineering is a
“wicked problem”
• Cannot be easily defined so that all stakeholders agree on
the problem to solve
• Require complex judgments about the level of abstraction
at which to define the problem
• Have no clear stopping rules
• Have better or worse solutions, not right and wrong ones
• Have no objective measure of success
• Require iteration — every trial counts
• Have no given alternative solutions — these must be
discovered
• Often have strong moral, political or professional
dimensions

S. Buckingham Shum
http://kmi.open.ac.uk/people/sbs/org-knowledge/aikm97/sbs-paper2.html
 Other problems

• Lack of well-understood
representations of software [Brooks]
makes customer and engineer
interactions hard
• Relatively young field
• Software intangibility is deceptive
 Law XXIII, Norman Augustine
[Wulf]

“Software is like entropy. It is difficult

to grasp, weighs nothing, and obeys
the second law of thermodynamics;
i.e., it always increases.”
 Dominant discipline
As the size of the software system grows, the key
discipline changes
[Stu Feldman, thru 107]

Code Size Discipline

103 Mathematics
104 Science
105 Engineering
106 Social Science
107 Politics
108 ??
Notkin’s Top 10 Observations
• About software engineering
– With apologies and appreciation to many
unnamed souls
• I’d appreciate help revising this list over the
quarter
• And, again, the intent of this is to convey,
now, many of my prejudices
– You’re not required to share them, but you’ll
understand more because I’m being explicit
about (most of) them
 1. Don’t assume similarity
among software systems
• Does (and should) the reliability of a nuclear
power plant shutdown system tell us much about
the reliability of an educational game program?
• Does (and should) the design of a sorting
algorithm tell us much about the design of an
event-based GUI?
• So, assume differences until proven otherwise:
not doing so causes a tremendous amount of
confusion in the degree of applicability of
different research approaches, tools, etc.
2. Intellectual tools dominate
software tools in importance
• How you think is more important than the
notations, tools, etc. that you use
• Ex: Information hiding is a key design
principle
– Interface mechanisms can enforce information
hiding decisions but cannot help one make the
decisions
• Ex: The notion of design patterns is more
important than languages that let you encode
them
 3. Analogies to “real”
engineering are fun but risky
• One reason is because of the incredible
rate of change in hardware and software
technology
– Wulf: what if the melting point of iron changed
by a factor of two every 18 months?
• Another is that software seems to be
constrained by few physical laws
• But I’ll make them anyway, I’m sure
– And you will, too
 Aside: should software
engineers be licensed?
• You may have heard about this issue
– For example, Texas now requires (under some
conditions) that software engineers be licensed as
professional engineers
• It’s an incredibly complex issue
– Technically, socially, politically and legally
– I’d be happy to discuss my views on this with
individuals (including on the mailing list), but I won’t
spend time in class on it
• BTW, I am strongly opposed to licensing
software engineers for the foreseeable future
4. Estimating benefits is easier
than estimating costs
• “If only everyone only built software my
way, it'd be great” is a common
misrepresentation
– Ex: The formal methods community is just
starting to understand this
• But at the same time, estimating the
costs and the benefits is extremely hard,
leaving us without a good way to figure
out what to do
 5. Programming languages
ensure properties distant from
the ones we want
• Programming languages can help a lot,
but they can't solve the "software
engineering" problem
• Ex: Contravariant type checking (such
as in ML) has significant benefits, but
regardless, it doesn’t eliminate all errors
in ML programs
– And covariant typing, with its flaws, may be
useful in some situations
 6. The total software lifecyle
cost will always be 100%
• Software development
and maintenance will
always cost too much
• Software managers will
always bitch and moan
• Software engineering
researchers will always
have jobs
• Software engineers will
always have jobs
 7. Software engineering is
engineering
• Although software engineering draws heavily
on mathematics, cognitive psychology,
management, etc., it is engineering in the sense
that we produce things that people use
– It’s not mathematics, nor cognitive psychology, nor
management (nor etc.)
• Nor logical poetry (cf. the Michael Jackson video we’ll
see later in the quarter)
• If somebody is talking about engineering
software without ever mentioning “software”,
run away
8. Tradeoffs are key, but we’re
not very good at them
• Getting something for nothing is great, but it isn't
usually possible
• We almost always choose in favor of hard criteria
(e.g., performance) over soft criteria (e.g.,
extensibility)
– This makes sense, both practically and theoretically
– Brooks’ Golden Rule doesn’t really work
– But the situation leaves us up a creek a lot of the time
• Maybe we’re about to get better at this as the cost
of people continues to grow
– But I doubt it
9. It’s good to (re)read anything
by Brooks, Jackson & Parnas
• “A classic is something everyone
wants to have read, but nobody wants
to read.” [Mark Twain]
• It’s more important to read their works
than to read the latest glossy rag or
modern book on the latest fad
• Really
10. Researcher  Practitioner
• Software engineering researchers
should have a bit of the practitioner in
them, and software engineering
practitioners should have a bit of the
researcher in them
• At the end of the quarter, I hope that I’ll
have more understanding of practice,
and you’ll have more understanding of
the research world
 Overview—five topics
• Requirements and specification
• Design
• Evolution (maintenance, reverse engineering,
reengineering)
• Analyses and tools (static and dynamic)
• Quality assurance and testing

• Yes, there is some overlap

• I reserve the right to completely change my mind about
the order and exactly what is covered!
 What’s omitted? Lots
• Metrics and measurement
– Some in QA
• CASE
– Some in evolution and tools
• Software process
– CMM, ISO 9000, etc.
• Specific methodologies
• [UX]ML
• Software engineering for specific domains (real-time,
the web, etc.)
• What else?
Requirements & specification
(2 lectures)
• Formal methods
– State-based, algebraic, model-
based
– Model checking
• Problem and domain analysis
– Problem frames, use-case,
collaborations, etc.
• Highlight: A Michael Jackson
video
 Design (2 lectures)
• Classic topics
– Information hiding
– Layered systems
– Event-based designs (implicit invocation)
• Neo-modern design
– Limitations of classic information hiding
– Design patterns
– Software architecture
– Frameworks
 Evolution (2 lectures)

• Why software must change

• How and why software structure
degrades
• Approaches to reducing structural
degradation
• Problem-program mapping
• Program understanding,
comprehension, summarization
Analyses and Tools (2 lectures)

• Static analyses
– Type checkers
– Extended type checkers
• Dynamic analyses
– Profiling
– Memory tools
– Inferring invariants
 Quality assurance (1 lecture)

• Verification vs. validation

• Testing
– White box, black box, etc.
• Reliability
• Safety (maybe)
Anything else?
 Overview of course work
• Four assignments, each of a different form
– A standard homework, a paper distilling research in an
area, an assessment of a research prototype tool, etc.
– All turned in electronically; each worth 23% of the grade
• The other 8% will represent your interaction in lecture
and (more importantly) on the mailing list
– Discussion of papers, of lectures, and of other software
engineering issues on your mind
• This is especially important for a distance learning class
• It’s the best way to learn from each other
– You are responsible for pushing the discussion threads,
although the TA and I will participate
 Grading: Let’s make a deal

• If you focus on the material and

don’t get compulsive about
grading …
• … then I will focus on the
material and not get compulsive
about grades
 Goodnight

• And don’t forget to buy those

course packs