Multimodal modeling, analysis and validation of op

1Multi-ModalModeling,Analysis,andValidationofOpenSourceSoftwareDevelopmentProcessesWaltScacchi1,ChrisJensen1,JohnNoll1,2,andMargaretElliott11InstituteforSoftwareResearchUniversityofCalifornia,IrvineIrvine,CA,USA92697-34252SantaClaraUniversitySantaClara,[e-mail],Genova,Italy,1-8,July2005.ReceivedBestPaperAwardAbstractUnderstandingthecontext,structure,activities,andcontentofsoftwaredevelopmentprocessesfoundinpracticehasbeenandremainsachallengingproblem.Intheworldoffree/opensourcesoftwaredevelopment,discoveringandunderstandingwhatprocessesareusedinparticularprojectsisimportantindetermininghowtheyaresimilartoordifferentfromthoseadvocatedbythesoftwareengineeringcommunity.PriorstudieshaverevealedthatdevelopmentprocessesinF/OSSDprojectsaredifferentinanumberofways.Inthispaper,wedescribehowavarietyofmodelingperspectivesandtechniquesareusedtoelicit,analyze,andvalidatesoftwaredevelopmentprocessesfoundinF/OSSDprojects,withexamplesdrawnfromstudiesofthesoftwarerequirementsprocessfoundintheNetBeans.orgproject.Keywords:Softwareprocess,processmodeling,opensourcesoftwaredevelopment,requirementsprocesses,empiricalstudiesofsoftwareengineering1.IntroductionIntheworldofgloballydispersed,free/opensourcesoftwaredevelopment(F/OSSD),discoveringandunderstandingwhatprocessesareusedinparticularprojectsisimportantindetermininghowtheyaresimilartoordifferentfromthoseadvocatedbythesoftwareengineeringcommunity.Forexample,inourstudiesofsoftwarerequirementsengineeringprocessesinF/OSSDprojectsacrossdomainslikeInternetinfrastructure,astrophysics,networkedcomputergames,andsoftwaredesignsystems[25,26,27],wegenerallyfindtherearenoexplicitsoftwarerequirementsspecificationsordocuments.However,wereadilyfindnumerousexamplesofsustained,2successful,andapparentlyhigh-qualityF/OSSsystemsbeingdeployedonaworld-widebasis.Thus,theprocessofsoftwarerequirementsengineeringinF/OSSDprojectsmustbedifferentthatthestandardmodelofrequirementselicitation,specification,modeling,analysis,communication,andmanagement[22].Butiftheprocessisdifferent,howisitdifferent,ormoredirectly,howcanwebestobserveanddiscoverthecontext,structure,activities,andcontentsoftwarerequirementsprocessesinF/OSSDprojects?Thisisthequestionaddressedhere.Ourapproachtoansweringthisquestionusesmulti-modalmodelingoftheobservedprocesses,artifacts,andotherevidencecomposedasanethnographichypermediathatprovidesasetofinformalandformalmodelsofthesoftwaredevelopmentprocessesweobserve,codify,anddocument.Why?First,ourresearchquestionspanstworealmsofactivityinsoftwareengineering,namely,softwaredevelopmentandsoftwareprocessmodeling.Sowewillneedtoaddressmultipleperspectivesorviewpoints,yetprovideatraceablebasisofevidenceandanalysisthatsupportsmodelvalidation.Second,giventherearealreadythousandsofself-declaredF/OSSDprojectsaffiliatedwithOSSportalslikeSourceForge.net,Freshmeat.netandSavannah.gnu.org,thenouranswerwillbeconstrainedandlimitedinscopetotheparticularF/OSSDproject(s)examined.ProducingamoregeneralizedmodeloftheF/OSSdevelopmentprocessbeingstudiedrequiresmultiple,comparativeprojectcasestudies,soourapproachshouldbecompatiblewithsuchagoal[25].Last,wewantanapproachtoprocessmodelingthatisopentoindependentanalysis,validation,communication,andevolution,yetbetraceabletothesourcedatamaterialsthatserveasevidenceofthediscoveredprocessintheF/OSSDprojectsexamined[cf.15].Accordingly,torevealhowweuseourproposedmulti-modelapproachtomodelrequirementsprocessesinF/OSSDprojects,wefirstreviewrelatedresearchtoprovidethefoundationalbasisforourapproach.Second,wedescribeandprovideexamplesofthemodelingmodesweusetoelicitandanalyzetheprocessesunderstudy.Last,weexaminewhateachmodelingmodeisgoodfor,andwhatkindofanalysisandreasoningitsupports.2.RelatedResearchandApproachThereisgrowingrecognitionthatsoftwarerequirementsengineeringcaneffectivelyincorporatemulti-viewpoint[7,16,22]andethnographictechniques[22,31]foreliciting,analyzing,andvalidatingfunctionalandnon-functionalsoftwaresystemproductrequirements.However,itappearsthatmanyinthesoftwareengineeringcommunitytreat3theprocessofrequirementsengineeringastransparentandprescriptive,thoughperhapsdifficulttopracticesuccessfully.However,wedonotknowhowlargedistributedF/OSSDprojectsperformtheirdevelopmentprocesses[cf.3].InitialstudiesofrequirementsdevelopmentacrossmultipletypesofF/OSSDprojects[25,26]findthatOSSproductrequirementsarecontinuouslyemerging[8,9,30]andassertedaftertheyhavebeenimplemented,ratherthanrelativelystableandelicitedbeforebeingimplemented.Similarly,thesefindingsrevealrequirementspracticecentersaboutreadingandwritingmanytypesofcommunicationsanddevelopmentartifactsas“informalisms”[25],aswellasaddressingnewkindsofnon-functionalrequirementslikeprojectcommunitydevelopment,freedomofexpressionandchoice,andeaseofinformationspacenavigation.Elsewhere,thereiswidespreadrecognitionthatF/OSSDprojectsdifferfromtheirtraditionalsoftwareengineeringcounterpartsinthatF/OSSDprojectsdonotingeneraloperateundertheconstraintsofbudg