No Silver Bullet: Essence and Accidents of Software ...hjemmesider.diku.dk/.../Slides/no-silver-bullet.pdf · No Silver Bullet: Essence and Accidents of Software Engineering by Frederick

No Silver Bullet

Adam HasselbalchHansen

Introduction

EssentialDifficulties

PastBreakthroughsHigh-level Languages

Time-sharing

Unified ProgrammingEnvironments

Hopes For TheSilverHigh-level andOO-Programming

Artificial Intelligence

Automatic and GraphicalProgramming

Program Verification

Environments and Tools

Promising Attackson the ConceptualEssenceBuy Vs. Build

Requirement Refinementand Prototyping

Great Designers

Summary

No Silver Bullet: Essence and Accidentsof Software Engineeringby Frederick P. Brooks Jr.

Adam Hasselbalch Hansen

Department of Computer ScienceUniversity of Copenhagen

Feb. 22, 2008

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

OutlineIntroduction

Essential Difficulties

Past BreakthroughsHigh-level LanguagesTime-sharingUnified Programming Environments

Hopes For The SilverHigh-level and OO-ProgrammingArtificial IntelligenceAutomatic and Graphical ProgrammingProgram VerificationEnvironments and Tools

Promising Attacks on the Conceptual EssenceBuy Vs. BuildRequirement Refinement and PrototypingGreat Designers

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

OutlineIntroduction





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

The What Now?

I A magic weapon to kill a mythical creature ofmysterious origin

I Software projects becoming monsters of misseddeadlines, blown budgets and flawed products

I Looking forward, no silver bullets can be seen.I No startling breakthroughs ahead, but there are

several smaller ones.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

The What Now?





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

The What Now?



I Looking forward, no silver bullets can be seen.

I No startling breakthroughs ahead, but there areseveral smaller ones.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

The What Now?





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

OutlineIntroduction





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Essential Difficulties I

I Software is not hardware

I No two-fold gains every 18 months — no Moore’sLaw

I No other technology with six O.o.M. performanceincrease in history. Ever.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Software is not hardwareI No two-fold gains every 18 months — no Moore’s

Law

I No other technology with six O.o.M. performanceincrease in history. Ever.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Software is not hardwareI No two-fold gains every 18 months — no Moore’s

LawI No other technology with six O.o.M. performance

increase in history. Ever.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Essential Difficulties II

Essence: The difficulties inherent in the nature ofsoftware

Accidents: Difficulties that are not inherent, butproblems nonetheless

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Essential Difficulties II

Essence: The difficulties inherent in the nature ofsoftware

Accidents: Difficulties that are not inherent, butproblems nonetheless

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Complexity

I Software is inherently complex

I Similar parts are made into subroutinesI Scaling software up (or down) does not linearly alter

complexityI Simplified “scale models”, as in math or physics,

won’t workI So managing software projects is complex, and

creates an understanding burden

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Complexity

I Software is inherently complexI Similar parts are made into subroutines

I Scaling software up (or down) does not linearly altercomplexity

I Simplified “scale models”, as in math or physics,won’t work

I So managing software projects is complex, andcreates an understanding burden

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Complexity

I Software is inherently complexI Similar parts are made into subroutinesI Scaling software up (or down) does not linearly alter

complexity

I Simplified “scale models”, as in math or physics,won’t work


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Complexity



won’t work


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Complexity



won’t workI So managing software projects is complex, and

creates an understanding burden

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Conformity

I Software must be “intuitive” and “easy to use”

I Different designers create difference interfacesI Confirmation from a human standpoint is one thing ...I ... but interoperability is a whole other!

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Conformity

I Software must be “intuitive” and “easy to use”I Different designers create difference interfaces

I Confirmation from a human standpoint is one thing ...I ... but interoperability is a whole other!

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Conformity

I Software must be “intuitive” and “easy to use”I Different designers create difference interfacesI Confirmation from a human standpoint is one thing ...

I ... but interoperability is a whole other!

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Conformity

I Software must be “intuitive” and “easy to use”I Different designers create difference interfacesI Confirmation from a human standpoint is one thing ...I ... but interoperability is a whole other!

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Changeability

I Unlike physical products (like a building), softwarecan be changed easily

I New uses for which no one designed the software,emerges

I New interoperability requirements emerges

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Changeability




No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Changeability




No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Invisibility

I No such thing as a “program blueprint”

I Many visualization representations existI but none gives a clear view.I Flow control, data control, dependency patterns,

data relationships, etc.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Invisibility

I No such thing as a “program blueprint”I Many visualization representations exist

I but none gives a clear view.I Flow control, data control, dependency patterns,


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Invisibility

I No such thing as a “program blueprint”I Many visualization representations existI but none gives a clear view.

I Flow control, data control, dependency patterns,data relationships, etc.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Invisibility

I No such thing as a “program blueprint”I Many visualization representations existI but none gives a clear view.I Flow control, data control, dependency patterns,


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

OutlineIntroduction





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

High-level Languages

I A factor five increase in productivity

I Eliminates unnecessary complexityI However, creates tool mastery burden

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I A factor five increase in productivityI Eliminates unnecessary complexity

I However, creates tool mastery burden

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I A factor five increase in productivityI Eliminates unnecessary complexityI However, creates tool mastery burden

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Time-sharing

I Preserves immediacy

I Essentialy: Code along while you compileI System response time shortened

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Time-sharing

I Preserves immediacyI Essentialy: Code along while you compile

I System response time shortened

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Time-sharing

I Preserves immediacyI Essentialy: Code along while you compileI System response time shortened

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Unified Programming Environments

I Integrated programming environments increaseproductivity

I Easier to use programs togetherI E.g. UNIX pipes, unified file formatsI Each tool applies to all others without extra work

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary



I Easier to use programs together

I E.g. UNIX pipes, unified file formatsI Each tool applies to all others without extra work

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary



I Easier to use programs togetherI E.g. UNIX pipes, unified file formats

I Each tool applies to all others without extra work

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary



I Easier to use programs togetherI E.g. UNIX pipes, unified file formatsI Each tool applies to all others without extra work

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

OutlineIntroduction





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

High-level and OOP

I Ada made a substantial difference in its day

I With OOP comes abstract and hierarchial data types(e.g. C++ Templates)

I But this requires the problem at hand to rely heavilyon these thing

I Only if nine tenths of a program is type specificationunderbrush will this have an O.o.M. gain.

I Which is doubtful...

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

High-level and OOP

I Ada made a substantial difference in its dayI With OOP comes abstract and hierarchial data types

(e.g. C++ Templates)

I But this requires the problem at hand to rely heavilyon these thing



No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

High-level and OOP


(e.g. C++ Templates)I But this requires the problem at hand to rely heavily

on these thing



No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

High-level and OOP



on these thingI Only if nine tenths of a program is type specification

underbrush will this have an O.o.M. gain.


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

High-level and OOP



on these thingI Only if nine tenths of a program is type specification

underbrush will this have an O.o.M. gain.I Which is doubtful...

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Using “intelligent programs” with generalizedinference engines

I Using a rule base, it yields conclusions and offersadvice

I The essential prerequisite for such a system is ateacher

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Automatic and Graphical Programming

Automatic programming: State the problem, and theprogram writes itself, based on that problem

I It is, however, the solution method, not the problem,that needs to be described.

Graphical Programming: Draw the program

I Programmers draw flowcharts after writing programsI Constraints in screen real estate

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary






No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary






No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary





I Programmers draw flowcharts after writing programs

I Constraints in screen real estate

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary






No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I A lot of effort goes into testing and reparing bugs

I Program verification does not mean error-proofprograms

I It’s relatively easy to verify a program. The hard partis making sure it’s properly designed to begin with.

I “Beware of bugs in the above code; I have onlyproved it correct, not tried it.” –Donald E. Knuth, 1977

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I A lot of effort goes into testing and reparing bugsI Program verification does not mean error-proof

programs

I It’s relatively easy to verify a program. The hard partis making sure it’s properly designed to begin with.


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary



programsI It’s relatively easy to verify a program. The hard part

is making sure it’s properly designed to begin with.


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary



programsI It’s relatively easy to verify a program. The hard part

is making sure it’s properly designed to begin with.I “Beware of bugs in the above code; I have only

proved it correct, not tried it.” –Donald E. Knuth, 1977

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Using proper tools to facilitate development

I Language specific editors, versioning systems, etc.I Giving developers a nice working environmentI Powerful workstations are not “the golden egg” (or

indeed the Silver Bullet)

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Using proper tools to facilitate developmentI Language specific editors, versioning systems, etc.

I Giving developers a nice working environmentI Powerful workstations are not “the golden egg” (or


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Using proper tools to facilitate developmentI Language specific editors, versioning systems, etc.I Giving developers a nice working environment

I Powerful workstations are not “the golden egg” (orindeed the Silver Bullet)

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I Using proper tools to facilitate developmentI Language specific editors, versioning systems, etc.I Giving developers a nice working environmentI Powerful workstations are not “the golden egg” (or


No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

OutlineIntroduction





No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Buy Vs. Build

I Before, business practices needed specializedsoftware

I Now, business practices adapt to the availablesoftware

I Generalized software, like spreadsheets, areextremely powerful and versatile

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Buy Vs. Build




No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Buy Vs. Build




No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Requirement Refinement and Prototyping

I The client does not know what he wants

I Prototype it!I Grow software. Don’t build it.I A working system, all the way though

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I The client does not know what he wantsI Prototype it!

I Grow software. Don’t build it.I A working system, all the way though

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I The client does not know what he wantsI Prototype it!I Grow software. Don’t build it.

I A working system, all the way though

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary


I The client does not know what he wantsI Prototype it!I Grow software. Don’t build it.I A working system, all the way though

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Great Designers

I Poor designs vs good designs can be caused byfaulty process

I Good designs vs great designs can notI Take care of designers. Give them ideal working

conditions. Travel funds, office size. Grow them.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Great Designers


I Good designs vs great designs can not

I Take care of designers. Give them ideal workingconditions. Travel funds, office size. Grow them.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Great Designers


I Good designs vs great designs can notI Take care of designers. Give them ideal working

conditions. Travel funds, office size. Grow them.

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Summary

I There are no silver bullets in the foreseable future

I However, combining several methods into a soundsoftware business strategy gets us a long way!

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Summary

I There are no silver bullets in the foreseable futureI However, combining several methods into a sound

software business strategy gets us a long way!

No Silver Bullet


Introduction



Time-sharing









Great Designers

Summary

Questions

?

Documents

No Silver Bullet: Essence and Accidents of Software ...hjemmesider.diku.dk/.../Slides/no-silver-bullet.pdf · No Silver Bullet: Essence and Accidents of Software Engineering by Frederick