Code Simplicity: Software Design In Open Source Projects

Max Kanat- lexander max!codesimplicity"com

#efore $%e S%ow

I tal& 'uic&ly Stop me if you don(t understand

lso stop me if you need examples )uestions or disagreements at end More detail on e,eryt%ing at my *log: %ttp:--www"codesimplicity"com-

Points will *e fast+ *ut still important

Some of t%ese t%ings you may already &now+ w%at matters is pointing t%em out as important"

.%o am I/

ssistant Project 0ead for #ug1illa .riter of Code Simplicity 0ong-time programmer and system administrator 2xperience programming and designing

Data comes from:

#ug1illa 3experiment4

Inter,iewing many programmers 5eading extensi,ely wit% analysis

Data collection is difficult *ecause of t%e timeframe of software projects"

I spea& ,ery definitely+ *ut please ma&e up your own mind a*out t%e t%ings I am saying"

.%at Is Software Design/

dministrati,e Decisions

.%at programmer to put w%ere De,elopment timeframes etc"

Coding Technical Decisions

.%at language to use .%at tec%nologies to c%oose etc"

$%ere Is 6o Science of Software Design

Science re'uires:

0aws Proof 5esults 3.aterfall4 3 gile4 etc"

Many met%odologies+ no science:

6ot going to pro,e anyt%ing today+ just s%ow

Can pro,e+ t%oug%+ in ,arious ways

$%ere are similar ideas out t%ere+ *ut t%ey are not t%e same as w%at I am going to tal& a*out

0argely t%ey are not low-le,el enoug%

Se,en Principles:
%ttp:--c7"com-cgi-wi&i/Se,enPrinciplesOfSoftwareDe,elopment

$%ey tell you what to do+ I only %elp you ma&e decisions for yourself and try to tell you why" I did not deri,e from any of t%ese met%odologies+ *ut t%e *its of t%em t%at wor& could *e deri,ed from w%at I am going to tell you"

.%y 8a,e a Science of Software Design/

8elp Ma&e $ec%nical Decisions Why do some t%ings 3wor&4 and ot%ers don(t/

5esults

#ug1illa Impro,ed My Own Programming

5esol,ed e,ery 'uestion

#roug%t no,ices to understand why 2xplained difficulties and 3war stories4 of experienced programmers

6ot #rainwas%ing or Mar&eting

6ot going to tell you w%at decisions to ma&e+ just going to gi,e you information t%at will %elp you ma&e t%em

$%is differs from met%odologies

#u11word-free

9OSS ,s" Proprietary

$%e *asics are t%e same+ *ut application can *e different"

Purpose Of Software

To help people

6e,er 3to %elp t%e computer4 Specific software is 3$o %elp people :*la%;4 S%ort Simple Specific 6eeded 9ollowed 2xactly

Stated purpose s%ould *e:

<oals of Software Design

#e as %elpful as possi*le Continue to *e as %elpful as possi*le Ma&e decisions t%at ma&e it easy to *e :and continue *eing; as %elpful as possi*le

Primary 0aw: 9uture

$%ere is more future t%an present

9uture is composed of infinite series of presents

$%e future is more important t%an t%e present 2ffort spent on design s%ould *e proportional to %ow muc% future time t%ere is in w%ic% you expect t%e software to exist

Planning to re-write is unnecessary

9uture: Known ,s" =n&nown

9ar 9uture 6ear 9uture 9uture 5e'uirements Conse'uences

Software %as long time lines

0aw Of C%ange

$%e longer your software exists+ t%e more pro*a*le it is t%at any piece of it will %a,e to c%ange

Means t%at as time goes on+ every piece is li&ely to c%ange

0aw of Defect Pro*a*ility

$%e c%ance of introducing a defect into your program is directly proportional to t%e si1e of t%e c%anges you ma&e

Perfection is impossi*le

.rite as little code as possi*le Don(t fix w%at isn(t *ro&en 2xplains re-use

0aw Of Simplicity

$%e ease of maintenance of any piece of software is directly proportional to t%e simplicity of t%e indi,idual pieces

6ot of t%e w%ole system+ just t%e indi,idual pieces

Stated differently+ is in,erse to t%e complexity" Simplicity is relati,e+ largely to ,iewpoint 8ow simple do you %a,e to *e/

Perspecti,e of anot%er programmer w%o(s ne,er seen your code

#e consistent

.%at Is a #ug/

Programmer(s Intentions

=ncertainties can *e resol,ed *y:

Comments Spec 35easona*le programmer4 ssume %e-s%e meant to do w%at is *est for t%e user

=ser 2xpectations

Specs t%at ,iolate user expectations are spec *ugs If t%ere(s a conflict+ it(s 3majority rules4

>ou can also add a preference+ *ut t%at adds complexity

.%ere Do #ugs Come 9rom/

Complexity

$%e *ox wit% a million unla*eled *uttons Particularly of language words+ sym*ols+ functions+ etc"

Misunderstandings

30aw4 of $esting

>ou don(t &now it wor&s unless you(,e tried it"

pplication in 9OSS

Must be more hardcore

0argely pro*lems wit% geograp%ic distri*ution

8ere(s w%ere I tell you some t%ings to do+ *ut you s%ould still ma&e up your own mind"

Difficulties of Design in 9OSS

Speed of c%ange can *e limited

5e,iews C%ec&in Procedures 0ac& of Some*ody to $al& $o :I5C %elps; Designer(s $ime

$ime a,aila*le can *e limited

8a,e to communicate designs 'uic&ly 8a,e to *e a*le to read design 'uic&ly

Implementer(s $ime

Communication *andwidt% is limited

8a,e to type to communicate design 6o w%ite*oard+ etc" <roup pro*a*ly won(t all *e t%ere at once 2it%er to de,elopment in general or just your project"

6o,ices

Desire for consistency ,s" desire for de,elopment speed Disconnection wit% users =ser re'uirements ? My re'uirements/

De,elopers pic& w%at to wor& on

May not want to conform to design May not want to wor& on refactoring

<etting out releases for testing

Solutions

2xtreme consistency

If you can(t communicate a design+ it %elps if t%e existing code already wor&s t%e rig%t way

#rief communications Code re,iews 2xtensi,e de,eloper documentation 0ots of attention to new*ies

#e nice

5ead support mail 5ead *logs a*out your product

#ut don(t e,er let your detractors write your re'uirements"

Sur,ey your users 8a,e some*ody w%o lo,es refactoring

$%e 2nd: ) @

%ttp:--www"codesimplicity"commax!codesimplicity"com