How open is open enough

来源：微智科技网

ResearchPolicy32(2003)1259–1285

Howopenisopenenough?

Meldingproprietaryandopensourceplatformstrategies

JoelWest∗

CollegeofBusiness,SanJoséStateUniversity,OneWashingtonSquare,SanJose,CA95192-0070,USA

Abstract

Computerplatformsprovideanintegratedarchitectureofhardwareandsoftwarestandardsasabasisfordevelopingcomplementaryassets.Themostsuccessfulplatformswereownedbyproprietarysponsorsthatcontrolledplatformevolutionandappropriatedassociatedrewards.

RespondingtotheInternetandopensourcesystems,threetraditionalvendorsofproprietaryplatformsexperimentedwithhybridstrategieswhichattemptedtocombinetheadvantagesofopensourcesoftwarewhileretainingcontrolanddifferentiation.Suchhybridstandardsstrategiesreﬂectthecompetingimperativesforadoptionandappropriability,andsuggesttheconditionsunderwhichsuchstrategiesmaybepreferabletoeitherthepurelyopenorpurelyproprietaryalternatives.©2003ElsevierScienceB.V.Allrightsreserved.

Keywords:Opensource;Standardscompetition;Computerarchitecture;Innovationreturns

1.Introduction

Theevolutionofthecomputerindustryhasbeendrivenbytheemergenceofstandardizedplatformsthatallowmodularsubstitutionofcomplementaryassetssuchassoftwareandperipheralhardware.

Theinitialplatformswereproprietary,inwhichacomputersystemsmanufacturercontrolledallhard-wareandsoftwarelayersofthestandardsarchitec-ture.Theseplatformswerelaterchallengedbytwohardware-independentoperatingsystems—UnixandWindows—whichreduceddifferentiationbetweenhardwarevendorsandshiftedplatformcontroltotheoperatingsystemvendors.TheUnixoperatingsystemalsoinspiredamoreradicalshift,theopensourcemovementinwhichLinuxallowedusersandcompetitorstocontrolaplatform’sdirection.

Tel.:+1-408-924-7069;fax:+1-408-924-3555.E-mailaddress:joel.west@sjsu.edu(J.West).

∗Thesevariousstrategiesreﬂecttheessentialtensionofdefactostandardscreation:thatbetweenappropri-abilityandadoption.Torecoupthecostsofdevelop-ingaplatform,itssponsormustbeabletoappropriateforitselfsomeportionoftheeconomicbeneﬁtsofthatplatform.Buttoobtainanyreturnsatall,thesponsormustgettheplatformadopted,whichrequiressharingtheeconomicreturnswithbuyersandothermembersofthevaluechain.

Theproprietaryandopensourcestrategiescorre-spondtothetwoextremesofthistrade-off.Inmakingaplatformstrategyforthe21stcentury,leadingcom-putervendorsfaceadilemmaofhowmuchisopenenoughtoattractenoughbuyerswhileretainingade-quatereturns.

FirstIreviewthetheoryandhistoryofproprietarycomputerplatformstrategies,contrastingthatwiththeUnix-basedopensystemsmovement.Ithenexam-inetheoriginsandmotivationsofLinuxandotheropensourcesoftwareprojects.Inthreeabbreviated

1260J.West/ResearchPolicy32(2003)1259–1285

casestudies,Ipresentthehybridstrategiesofthreeplatformvendors—AppleComputer,IBMandSunMicrosystems—thatcombineopensourceandpropri-etaryplatformstrategiesinhopesofobtainingcom-petitiveadvantage.Finally,ananalysisofthesecasesiscombinedwithpriorresearchtosuggestthetheo-reticalimplicationsofsuchhybridstrategies.2.Proprietaryplatformstrategies

2.1.DynamicsofproprietaryplatformcompetitionAproprietaryplatformconsistsofanarchitec-tureofrelatedstandards,controlledbyoneormoresponsoringﬁrms.Foracomputersystem,thearchi-tecturalstandardstypicallyencompassaprocessor,operatingsystem(OS),andassociatedperipherals.Somehavealsoextendedtheconceptofa“platform”toincludemultiplelayersofsoftware,suchasap-plicationsthatrelyona“middleware”toolsuchasJavaoradatabase(MorrisandFerguson,1993;BresnahanandGreenstein,1999;WestandDedrick,2000).

AplatformisbutaspeciﬁcexampleofthegeneralclassoftechnologicalinnovationsstudiedbyTeece(1986),wholinkstheabilityofﬁrmstoproﬁtfromtheirtechnologicalinnovationstotheappropriabilityregimeforintellectualpropertyrights(IPR)—eitherthroughformaldejureprotection(e.g.patents)orthroughdefactoprotectionsuchastacitknowledgeortradesecrets.AbsentsuchIPRprotection,ﬁrmssellingagiventechnologycanbeexpectedtoadoptmarginalcostpricinganddriveproﬁtmarginstozero(KatzandShapiro,1986;BeggsandKlemperer,1992).Withoutappropriability,Teece(1986)suggeststhatﬁrmsmustusesomecombinationofspeed,timingandluckiftheyhopetoappropriatereturnsgeneratedbytheirin-novation.

Teece(1986)alsoconsidersthoseinnovations(suchascomputersystems)thatrequiretheprovi-sionofcomplementaryassetstocommercializetheinnovation.Whenadditionalinvestmentisrequiredtoco-specializetheassettobeusefulwithagiveninno-vation,thesuccessfuladoptionoftheinnovationandtherelatedassetsaremutuallyreinforcing,providingapositivefeedbackcycle.Thus,tomakeasuccessful“wholeproduct”solution,theowneroftheinnovationseekstoattractsuchcomplementaryassets,inpartbysharingtheoverallreturnsoftheinnovationwiththethird-partysupplierofsuchassets(KatzandShapiro,1985;Teece,1986;Moore,1991).

Thepositivefeedback,self-reinforcingcycleofsuc-cessbetweenadefactostandardanditsco-specializedassetsuccesshasbeentermed“networkexternalities”(KatzandShapiro,1985)or“demandsideeconomiesofscale”(KatzandShapiro,1986).Whensuchnet-workeffectsarecoupledwithswitchingcostsbe-tweenstandardsandhighupfrontR&Dcosts,Arthur(1996)predictsthatthedominanttechnologywillenjoy“increasingreturnstoscale”thatmagnifyanearlyleadinatechnologycontest.Theinstabilityandself-reinforcingnatureofsuchaleadhasoftenbeenreferredtoas“tipping”ofthecontest(e.g.BesenandFarrell,1994).LiebowitzandMargolis(1999)arguethatactualtippingisrare,andthatlastingsuccessismoreoftenexplainedbyproductioneconomiesofscaleandﬁrmexecution.

Inthecaseofcomputingplatforms,mostresearchhasfocusedononeparticulartypeofcomplemen-taryasset,thatofprepackagedapplicationsoftware.Acomputerplatformisnot,initself,usefulwith-outsoftwaretosolvespeciﬁcproblems.Duringthe1960sand1970s,largeorganizationsbuyingmain-framecomputerstypicallydevelopedtheirowncustomsoftware.However,theadventofthemass-marketper-sonalcomputersattractedmanynewusersunabletodeveloptheirownsoftware,fuelingashifttoprepack-agedapplicationsoftwarepackages(Mowery,1996).Assuch,thecontroloftheplatform’scomplementaryassets(e.g.packagedsoftware)isdeterminedbytheabilitytocreateandevolveapplicationprogramminginterfaces(APIs),whichspecifyhowapplicationsoft-waremustbeco-specializedtoworkwithaparticularplatform(WestandDedrick,2000).

Multipleplatformscanandhavesimultane-ouslyco-existedservingdifferentmarketsegments(Table1).BresnahanandGreenstein(1999)arguethatintheUScomputerindustry,newplatformssuc-ceededwhentheytappedunservedmarketniches,avoidingcompetitionwithestablishedplatformsuntiltheyachievedcriticalmass.Duetosimpleeconomiesofscale,mass-marketplatformsdisplacedmorespe-cializedproducts,eitherbyprovidinglowercostoraddressingabroaderrangeofbuyerneeds(MorrisandFerguson,1993).

J.West/ResearchPolicy32(2003)1259–1285

1261

Table1

Major20thcenturycomputingplatformsCategoryFirmPlatformReleasedMainframe

IBM

IBMS/36019(S/370,S/390)MinicomputerDEC

VAX/VMS1977(OpenVMS)IBM

AS/4001988AT&T,laterOSFa

Unix

1980b

WorkstationApolloDomainc1980Sun

SunOS(Solaris)19828-bitPCApple

AppleIIc1977DigitalResearchaCP/Mc1976b≥16-bitPC

IBM

IBMPC1981MicrosoftaWindows1990bNECPC-98c1983AppleMacintosh1984PersonalPalm

Pilot1996digitalassistant

Microsofta

WindowsCE

1996

OSvendor;otherwise,vendorsareverticallyintegratedman-ufacturers.

bWidespreadcommercialrelease.cDiscontinued.

Firmsthatsuccessfullyestablishandmaintainaproprietaryplatformenjoytherighttoappropriatethereturnsfromasuccessofthatplatform(MorrisandFerguson,1993).Butplatformsuccessisnecessarybutnotsufﬁcientconditionforproﬁtingfromproprietarytechnologyinnovation.Whencompetingﬁrmscontroldifferentlayersofthestandardsarchitecture,platformleadershipisunstablebecausecontroloftheplatformcanshiftwithoutdisruptingthebuyer’svalueproposi-tion.Inparticular,aﬁrmthatcantakecontrolofaccesstothecomplementaryassetshasanincentivetodosotocapturethereturnsfromtheplatform(BresnahanandGreenstein,1999;WestandDedrick,2000).

Inexplainingplatformsuccess,economicresearchhasfocusedondemandandsupplysideeconomiesofscale,andthebroadstrategicchoicesmadebyspon-soringﬁrms.However,thereisalsorepeatedevidencethatoperationalexecutioniscrucialtotherelativesuc-cessorfailureofindividualplatforms(MorrisandFerguson,1993;LiebowitzandMargolis,1999;West,2003).Also,measuresofplatformsuccesshavefo-cusedonadoptionormarketshare.Themoremanage-riallyrelevantmetricwouldbethesponsor’snetproﬁt

orreturnoninvestmentfromtheproprietarytechno-logicalinnovation,althoughsuchdataismuchharderforresearcherstoobtain.

2.2.Mainframes:verticallyintegratedproprietaryplatforms

In19,IBMintroducedtheworld’sﬁrstsuccess-fulcomputingplatform,theSystem360(S/360).Akeysuccessfactorwasthemodulararchitecturethatenabledtheuseofthesamesoftwareandperipheralsthroughouttheproductline,providinginteroperabilitythatwasmissingtheproductlinesofIBMandothercompanies.IBMalsoleverageditsexistingdomes-ticmarketshareandglobalreachtowinthelargestshareoftheglobalmarket,vanquishingtheproprietaryplatformsofdomesticrivalsandEuropeannationalchampions(Flamm,1988;Chandler,1997;Moschella,1997).TheoneexceptiontoIBM’sdominancewasinJapan,wherebythe1990sIBMheldonlya25%shareaspartofastable,fourﬁrmoligopoly,inwhichIBM’srivalsproducedclonesoftheS/360platform(Anchordoguy,19;FergusonandMorris,1993).IBMwasaverticallyintegratedmanufacturerofprocessors,systems,peripheralsandsoftwarefortheS/360platformanditsSystem370(S/370)andSys-tem390(S/390)successors.ThepatternwasrepeatedbyDigitalEquipmentCorp.,whichintroducedase-riesofminicomputerplatforms,culminatingwithitsmostsuccessfuloffering,theVAX32-bitminicom-puter(1978)withitsproprietaryVAX-11processorandVAX/VMSoperatingsystem.DECdominatedUStechnicalmarkets,althoughIBMledbusinessmar-ketswithitsAS/400minicomputer(BresnahanandGreenstein,1999).

Lateentrantshadtroublegainingmarketsharefortheirinnovationsduetoswitchingcostsbetweentheproprietaryplatforms(Greenstein,1997).Amajorcostinconvertingfromoneplatformtoanotherwasduetoapplicationsoftware,whichhadtobespecial-izedtoﬁteachplatform’sAPIs.Inthe1960sand1970s,thistendedtobesoftwarecustomdevelopedtomeettheﬁrm’sparticularneeds,butsubsequentbuyershiftstousingoff-the-shelfpackagedsoftwareshiftedtheconversioncostfromuserstothird-partysoftwaresuppliers.

Chandler(1997)arguesthattheconcentrationoftheglobalmainframeindustrywasconsistentwiththe

1262J.West/ResearchPolicy32(2003)1259–1285

patternofothercapital-intensiveindustries,inthatthehighentrycostsallowthepioneerandearlychal-lengerstoformastableoligopoly.BresnahanandGreenstein(1999)attributesucholigopolytotheef-fectofendogenoussunkcostsinrewardingscale;theirmechanismsandpredictedoutcomescorrespondtoArthur’s1

(1996)formulationofincreasingreturnstoscale.2.3.Personalcomputerbringshorizontalplatformcontrol

Aswithmainframesandminicomputers,theper-sonalcomputerindustryattractedmanynewentrantsthattriedandfailedtoestablishsuccessfulplatforms.However,therewasonecrucialdifference:the1971inventionofthemicroprocessordramaticallyloweredthecostofentryandalsoledtoplatformconvergenceasalargenumberofsystemsmakerspurchasedpro-cessorsfromashrinkingnumberofmicroprocessorvendors.Duringtheinitial8-bitera,thePCindus-tryusedtwodifferentplatformstrategies.OnegroupusedtheIntel8080(orcompatible)processorsalongwithCP/M,aproprietaryoperatingsystemlicensedtomanycomputermakers.Theothergroupboughtaninexpensiveprocessorandthendesignedtheirownsoftwaretorunontopofit.ApplicationsoftwarewaseitherdesignedforCP/MAPIs,orforoneofthepro-prietaryplatforms.

SeeingthegrowthofthePCmarket,andworriedaboutcedingmarketcontroltoearlypioneers,in1980IBMlaunchedacrashprojecttobuilda16-bitPC.IBM’smainframepowerandreputationassuredthesuccessoftheIBMPC,whichalsorendered8-bitPCsobsolete.TheIBMPCstandardsoondominatedtheworld,exceptinJapanwhereNEC’sproprietaryPC-98dominatedthemarketfrom1983to1995(ChposkyandLeonsis,1988;WestandDedrick,2000).

AswithCP/M,IBM’sPCarchitectureusedbothaprocessorandoperatingsystemfromoutsideven-dors.WhencoupledwithitsunexpectedlegaldefeatsonROMcopyrights,IBMlostcontrolofitsplatform

TheBresnahanandGreensteinformulationalsoidentiﬁesplatform-speciﬁcinvestmentsinprovidingmarketstability,corre-spondingtoTeece’s(1986)earlieranalysisofspecializedcomple-mentaryassets.

asotherﬁrmsproduced“clone”computersthatranthesameapplicationsoftware(Langlois,1992).Forthenextdecade,IBMspentbillionsofdollarsonpro-prietarytechnologiesinanunsuccessfulattempttore-assertitsleadershipofthePCindustry.

Instead,the“IBMPC”platformgraduallybegantobetermedthe“Wintel”platform,anacknowledgmentthattheproprietaryplatformcontrolrestedwithMi-crosoft(Windows)andIntel.BothﬁrmsenjoyedthebarrierstoimitationprovidedbyeconomiesofscaleforR&Dandnetworkeffectsthroughsoftwaresupply(Arthur,1996).Grove(1996)arguesthattheresultanthorizontalspecializationofthePCindustryismoreefﬁcient(andthusmoredurable)thantheverticallyintegratedstructure,becauseitallowsfortheproducerofeachlayertoachieveeconomiesofscalebyservingthebroadestpossiblemarket.

2.4.Workstations:Unixandopensystems

Thehorizontallyspecializedplatformstrategypre-datesCP/MandMS-DOS,butinsteadbeganwithAT&T’sUnixoperatingsystem.Unixbeganin1969withminicomputers,helpedcreatethenewwork-stationplatformsofthe1980s,andlaterbecameanimportantmulti-vendor,multi-productmainframeplatform.Unixeventuallyevolvedintoanewformofopennon-proprietaryplatformstandard,oftenreferredtoasthe“opensystems”movement(Gabel,1987).Aswithanyotheroperatingsystem,Unixwasnotacompleteplatformspeciﬁcation,becauseeachhardwaresystemmighthavedifferentprocessorandperipheralinterfacestandards(Table2).However,UnixquicklyevolvedintoaportableOSthattendedto“hide”thedifferencesbetweenhardwarefromsoftwareapplications,andsocouldpresentasetofcommonAPIsacrosswidelydivergenthardwareimplementations.ThetaskwasaidedbytheCpro-gramminglanguagewhichservedasahighlyefﬁcientsubstituteforhardware-dependentassemblylanguage.AT&Twasrestrictedfromsellingcomputerprod-uctsbya1956anti-trustsettlement,sointheﬁrst15yearsUnixwaslimitedtointernaluse,researchuni-versities,andacomparativelysmallnumberofusercompaniesandhardwarevendorsthatboughtsourcecodelicenses.AT&Teventuallyspun-offUnixintoaseparatecompany,whichwassoldtoNovellandlaterSCO.Inthe1980s,Unixbecamethepreferred

J.West/ResearchPolicy32(2003)1259–1285

Table2

Ownershipofachitecturallayersforrepresentativecomputerplatforms,1990ProductVendorIBMDECDECSunAppleIBMCompaq

1263

Platform

ProductS/390

VaxStationVaxStationSparcStationMacintoshPS/2DeskPro

SegmentMainframeWorkstationWorkstationWorkstationPCPCPC

ApplicationsOwn,Own,Own,Own,Own,Own,Third

thirdpartythirdpartythirdpartythirdpartythirdpartythirdpartyparty

OS/390(own)VMS(own)

Unixa(thirdparty)Unixa(thirdparty)MacOS(own)

MS-DOS(thirdparty)MS-DOS(thirdparty)

CPU

ES9000(own)CVAX(own)CVAX(own)Sparc(own)

68030(thirdparty)386(thirdparty)386(thirdparty)

Licensedfromathird-partysupplierbutwithproprietaryextensions.

operatingsystemforcomputerworkstationsandalsowonsigniﬁcantmarketshareinminicomputers,high-endcomputerserversandsupercomputers.

Insomeways,AT&T’sUnixstrategiesinthe1980sparalleledthoseofMicrosoftwithMS-DOSandlaterWindows.Bylicensingtheiroperatingsystemstomul-tiplehardwarevendors,eachmadetheirplatformubiq-uitousbyreducingswitchingcostsanddifferentiationbetweenhardwarevendors.BothoperatingsystemssharedAPIsacrossmultiplehardwarevendors.Aswitheventhemostproprietarycomputingplatform,UNIXandMS-DOSwere“open”tothird-partysoft-waresuppliers,utilizingAPIswidelydisseminatedtomaximizesoftwareavailability.

UnixandWindowswerealsosimilarinthatthesharedOSnearlyeliminatedtheantecedentsofArthur’s(1996)typology—economiesofscale,net-workeffectsandswitchingcosts—thatmightleadtopositivereturnstoscale,andthustipthemarketsharecontesttoasinglewinneramongcomputermanufac-turers.2Thelackofsuchfactorsmadeitunlikelythatanymanufacturerwouldenjoymarketdominance.InUnixworkstations,mostofthemarketwasfrag-mentedamongfourmajorﬁrms—Sun,IBM,HPand

ForbothUnixandWindowsmachines,thecomputermakersspentlittleonoperatingsystemresearchanddevelopment,elimi-natingthatasasourceofeconomiesofscale.ThesharedAPIsalsoprovidedaccesstothesamecomplementaryassetsandreducedswitchingcostsbetweencomputermakers.Between“Wintel”ma-chines,thecomplementaryassetswereidenticalandswitchingcostsnegligible;forUnixworkstations,theprocessordifferencesandproprietaryAPIextensionsslightlyincreasedtheswitchingcostsforusersandsuppliersofapplicationsoftware,butweremuchlowerthanbetweenproprietaryplatforms.

DEC(laterCompaq)—noneofwhomcapturedmorethana35%marketshare.

However,inthe1990sUnixandWindowsdiffereddramaticallyinthecontroloftheirAPIs.Windowsre-taineditsproprietaryAPIsunderthecontrolofasingleﬁrmwhichproducedtheonlyimplementation,allow-ingittoenjoymonopolyrents.Meanwhile,defactocontroloftheUnixAPIshadshiftedtovariousin-dustrycommitteesandconsortiainthe“opensystemsmovement”,whichpublishedvendor-independentstandardssuchasPOSIX,OSF/1andX/OPEN.Whilethetrademarked“Unix”wasallderivedfromAT&T’sIPR,the“opensystems”evolvedintomultipleinde-pendentimplementations—includingseveral“opensource”implementations—eachcompliantwiththeacceptedPOSIXspeciﬁcation.Thus,theopensystemsmovementreﬂectedanevolutionofplatformstrate-giesthatreducedtheabilityofanyindividualﬁrmtoobtaincontrolordifferentiationfortheirplatform(Table3).However,thesubsequent“opensource”movementtookthistothenextlevel.2.5.Assessment

AfterIBMsucceededwithitsS/360platform,rivalcomputermakerssoughttoemulateIBM’sverticallyintegratedproprietaryplatformstrategy.Positivere-turnstoscalemeantthewinningproprietaryplatformenjoyedhighbarrierstoimitationandthushighprof-its.Verticalintegrationallowedaﬁrmtoappropriatethoseproﬁtswithouthavingtosharethemwithotherﬁrms.Butwhenasingleplatform(liketheIBMS/360)enjoyedsustainedmarketsharedominance,ri-valshadtroublecompetingwithaverticallyintegrated

Table3

RepresentativeplatformIPRstrategies,1990Platformstrategy

Sponsor

J.West/ResearchPolicy32(2003)1259–1285

MultiplehardwarevendorsMultipleOSvendors

Sourcelicensing

PublishedAPIsforthird-partysoftwareYesYesYesYes

APIcontrolProducts

Mostopen

Verticallyintegratedproprietary

HorizontalproprietaryUnixMostclosedOpensystems

HardwarevendorsMicrosoftAT&TConsortia

NoYesYesYes

NoNoNoYes

NoNoYesYes

HardwarevendorSoftwarevendorSoftwarevendorConsortium

IBMS/360,DECVAX,MacintoshWindows3.0SystemVOSF,X/Open

proprietarystrategy:withasmallershare,theylackedminimumefﬁcientscaletocovertheﬁxedR&Dcosts.

Asaninterimmeasure,ﬁrmsprocurednon-criticalcomponents(suchasmemoryandperipherals)fromcommonsuppliers,developingaproprietaryproces-sorandoperatingsystemtoprovidedifferentiation.However,suchcomponentsourcingdidnotaddresstheprocessorandOSR&Dcosts,andstillleftmarketsharelaggardsatasigniﬁcantdisadvantageduetofewercomplementaryassets(typicallysoft-ware)andswitchingcostsfacedbymostpotentialusers.

Sooverthelongerterm,minorcomputermakerstriedvariousstrategiestopoolR&Dandthesupplyofcomplementaryassetsacrossmultipleproducers;thesestrategiesalsoallowedﬁrmswithoutproprietaryplatformstoenternewmarkets.OnesuchstrategywastobuythecrucialOSorprocessorfromexternalpro-prietaryvendors(suchasAT&T,MicrosoftorIntel).Anotherwastoformamulti-vendorconsortia(e.g.theOpenSoftwareFoundation)topooltechnologyamongvendors.Bysharingsome(orall)oftheirplatform,theyenjoyedbetteradoptionbutriskedintra-platformcompetitionthatlimitedtheirabilitytoproﬁtfromtheplatformsuccess,evenforﬁrms(likeIBM)thathadbeentheplatform’soriginalinnovator.

Whilethesenewerplatformstrategiesgavecom-puteruserslowerswitchingcostsandhigherbar-gainingpower,buyersstillfundamentallylicensedtechnologythatwasownedbycomputervendors(ortheiralliances)—ownersthatcouldsetthetermsandpricingofthetechnology,aswellastheschedulefor

enhancementsanderrorcorrection.3Thisrelationshipfundamentallychangedwiththeemergenceof“opensource”computingplatforms.

3.Emergenceofopensourceplatforms

Indevelopedcountries,softwareenjoysstrongintel-lectualpropertyrightsprotectionsintheformoftradesecrets,copyrightsand(mostrecently)patents.Theabilitytocreateandmodifysoftwareproductsisgov-ernedbytheaccesstothesourcecode,whichiswhyfor-proﬁtsoftwaredevelopershavehistoricallytreatedsuchsourcecodeasacloselyguardedtradesecret.So-called“opensource”softwarerepresentsthean-tithesesofaproprietarytechnologystrategy.RatherthanusingformalIPRprotectiontosetboundariesbe-tweenvendorsandtheircompetitorsandcustomers,opensourceenlistsallascollaborators,maximizingadoptionthroughoutthevaluechainbutminimizingtheoptionsforappropriatingrentsfromthesoftware.Thesuccessoftheopensourcemovementreﬂectedaconﬂuenceofthreefactorsinthemid-1990s(WestandDedrick,2001):

•usersseekinganinexpensiveUniximplementationfreeofAT&Tlicensingrestrictions;

•aphilosophicalmovementrejectingtheideaofsoft-wareownershipandappropriability;

Theopensystemsstandards-settingconsortiaincludedrepre-sentativesofmajorcomputerbuyers,butthedeploymentofnewtechnologiesimplementingthesestandardsremainedunderthecontrolofthecomputermakers.

J.West/ResearchPolicy32(2003)1259–12851265

•emergenceoftheInternetasbothanenablerandobjectiveforcollaborativesoftwaredevelopment.3.1.LinuxandotherUnix-likeplatforms

Inthe1980s,theUnixplatformhadthreemainat-tractionsforprogrammers:itranoninexpensivemini-computers,washardwareindependentandprovidedastateoftheartenvironmentforsoftwaredevelopment.Thus,itwasanaturaltargetforthosewhoeventuallydeveloped“clone”operatingsystems.

In1984,RichardStallmanlefthisjobasanMITprogrammertodevelopafreeUnix-likeoperatingsys-tem.HefoundedProjectGNU,whichby1990hadproducedavarietyofsoftwaredevelopmenttools,butlackedtheessentialcoreofamodernoperatingsys-tem,akernel(Stallman,1999).Effortstodevelopakernelhadﬁrstbeendelayedbyworkonothercom-ponents,andthenﬂounderedforseveralyearswithoutstrongleadership.

TwogroupsusedtheGNUtoolsasthebasisforassemblingcompleteandfreeUnix“clone”operat-ingsystems.In1991,LinusTorvaldsbeganwritingaUNIX-compatibleoperatingsystemforhisnewPC,andsolicitedotherstojoininhisefforts.Byearly1993,aversionofLinuxwasfreelyavailablefordownloadingontheInternet(Varhol,1994;TorvaldsandDiamond,2001).

Meanwhile,from1990to1992aUnixteamattheUniversityofCalifornia,Berkeleysolicitedout-siderstovolunteertorewritecomponentsforitsBSDUnixusingonlythepublishedAPIs(andthusnotviolatingthecopyrightandtradesecretsofAT&T’ssourcecode).ThisinturnspawnedaseriesofBSDimplementationsprimarilyaimedatPChardware:NetBSD(1993),FreeBSD(1993)andOpenBSD(1996)(McKusick,1999;WestandDedrick,2001).ByrelyingheavilyupontheProjectGNUcompo-nents,boththeLinuxandvariousBSDteamsdeliveredfreeUnix-compatibleoperatingsystemsoriginallyori-entedatpersonalcomputerhobbyists.Throughcostandﬂexibilityadvantages,theygraduallysupplantedotherUnixdistributionsforIntel-basedpersonalcom-puters,andalsoenabledsuchPCstobeusedasreliableserversbyorganizations.Despitecontrastingstrate-giesforcontrolandIPR,boththeLinuxandBSDgroupsbecameforerunnersofwhatlaterwasnamedthe“opensource”movement.

3.2.“Freesoftware”versus“opensource”Inhisacademiccomputingcareer,Stallmancametoexpectacomputingenvironmentwhereuserssharedsoftwareandcouldmakecustommodiﬁca-tions(Moore,2001).Suchsharinghadalsoexistedbetweensomelargecomputersites,throughcom-puterusergroupslikeSHARE(forIBMusers)andDECUS(forDigitalEquipmentCorp.users),andinafewpubliclydistributedsoftwareprogramssuchastheTENEXoperatingsystemandthesendmailmailserver.InlaunchingProjectGNU,Stallmanpromotedaphilosophydiametricallyopposedtothenormsofproprietarycommercialsoftware.Extendingbyanal-ogythetraditionsofpooledscientiﬁcresearchandthefreedisseminationofideas,Stallmanarguedthatallsoftwareshouldbe“freesoftware”,withsourcecodethatcanberead,modiﬁedandredistributed(Zachary,1991;Stallman,1999).

SomebutnotallofStallman’sgoalsweresharedbyotherdevelopersoffreesoftware.By1998,LinuxandothertechnologieshadbecomepopularbutStallman’sideologyhadwononlylimitedcommer-cialsupport.Topromoteadoptionbybusinessusersandthird-partydevelopers,ﬁrmsthatsoldLinux-andGNU-relatedsupportandservicesmettopromoteamorebusiness-friendlyconceptofcollaborativesoft-waredevelopment.Theylabeledtheircommonvision“opensource”(DiBonaetal.,1999,p.3).

Unlikeforproprietarysoftware,forboth“opensource”and“freesoftware”thesourcecodeandexecutablecomponentarefreelydistributed.BothcanbefreelydownloadedfromtheInternet,eventhoughsomeusersdecidetobuyadistributionontapeorCD-ROM.Andinbothcases,consulting,sup-portandtrainingservicescanandaresoldwithoutrestriction.

Themajordifferenceisthat“freesoftware”pro-hibitsexpostappropriationofthetechnology:anyderivativeworksmustalsobedistributedas“freesoftware”andallchangesreturnedtotheoriginalauthorforsubsequentredistribution.Stallman(1999)arguedthatthisisessentialtopreventﬁrmsfrommakingminorimprovementstofreesoftwareandthenusingitasawaytoattractuserstotheirnon-freecommercialupgrade.TheProjectGNUtoolsandtheLinuxOSweredistributedundertherestrictionsoftheso-calledGNUPublicLicense(GPL).

1266J.West/ResearchPolicy32(2003)1259–1285

Meanwhile,“opensource”projectsdidnotimposeanysuchrestrictions,allowingindividualsorﬁrmstocustomizeandcombineopensourcesoftwareastheydesired;however,theproﬁtpotentialofminorim-provementsislimitedbytheavailabilityofthefreealternative.Thisformofsourcecodelicensewasde-velopedbytheBSDUnixclonesandtheApachewebserver,andwaslateremulatedbyotherprojectsthatused“BSD-style”or“Apache-style”licenses(WestandDedrick,2001).4

Thetwolicensesdifferintheircompetitiveim-plications.AdeveloperreleasingsourcecodewithaBSD-stylelicensegrantsthemostrights,inthatothers(includingcompetitors)canmodifyandusethesoft-wareastheyplease.UsingtheGPLlevelstheplayingﬁeld:allusersarerequiredtoshareanysubsequentchanges,eliminatingtheabilityofanyparty(includ-ingtheoriginalownerandanycompetitors)todiffer-entiatetheirofferingsthroughsoftwareenhancements.Theyaresimilarinthatforbothlicenses,asoftwaredevelopervoluntarilysurrenderstheabilitytoappro-priatethereturnsfromitsR&Dinhopesofwinninggreateradoption.Althoughtheycannotdirectlyproﬁtfromsuchsoftware,developerscouldusefreesoft-wareasacomplementaryassettohelpsellhardware(orothersoftware),orcoulduseittosellassetscom-plementarytothefreesoftware—suchasconsulting,supportandtraining.4.Contextforthestudy

Thisstudyexaminesthedecisionbyproprietaryplatformvendorstoreleaseandsupportopensourcetechnologiesaspartoftheirplatformstrategiesintheperiod1995–2002(Table4).Theﬁeldstudyandanal-ysisofsecondarydatawereintendedtoexplainthissomewhatparadoxicaldevelopment.Wasitatempo-raryphenomenontiedtoaparticulartimeandplace,e.g.theimportanceofsupportingInternetstandards?Wasitadesperationphenomenonforﬁrmsthatwereeventuallydestinedtofail—ashappenedduringtheearly1990swhenfailingminicomputermakersaban-donedtheirproprietaryOStoadoptopensystems?Orwasthisasustainableandongoingbusinessmodelfor

Henceforward,Iuse“opensource”tosubsumeboththeBSD/Apache-stylelicensesandGNU-style“freesoftware”li-censes.

makersofcomputerhardwareandintegratedcomput-ingplatforms?

Thestudyconsidersthreecompanies—Apple,IBMandSun—thathistoricallyhadpromotedverticallyintegratedplatformsdifferentiatedbyproprietarysoftwarein(respectively)thePC,mainframeandworkstationeras.Astheseproprietarystrategiesbe-gantofalter,eacheventuallyemployedopensourcetorevitalizetheirstrategies.Atthebeginningofthestudy,thereweretwoobvioussimilaritiesbetweenthethreecompanies:allthreefacedseriouscompetitivepressuresfromMicrosoft,andallthreefacedachal-lengeinformulating“open”strategiesthatnonethe-lessallowedthemtoretainoneormoresourcesofcompetitiveadvantage.

4.1.RespondingtotheMicrosoftchallengeTherealityinthelate1990swasthatasinglecom-panydominatedtheITindustry:Microsoft.Whileitsharedthe“Wintel”platformcontrolwithIntel,unlikeIntelMicrosoftdidnothaveacompetitorinsupplyingPCmanufacturers.MicrosoftalsoplayedamajorroleinPCapplicationsoftware,serveroperatingsystems,serverapplicationsandanincreasingroleinmobiledevices.

IBM,AppleandSunsoughtnewstrategiestore-spondtopressurefromMicrosoft.WithitsPC,IBMhadrelinquishedindustryleadershiptoMicrosoft;ApplehadtraditionallybeenMicrosoft’sprimarycompetitorin16-bitPCplatforms;andSunfounditscoreworkstationandserverbusinessthreatenedbyMicrosoft’sattemptstoexpandbeyondthePC.

ManyhardwarevendorshadbeneﬁtedfromMicrosoft’stechnicalleadership.FirmssuchasNCR,Siemens,andUnisyssoldcomputersbuiltoncom-moditycomponents,focusingonaparticulargeo-graphicormarketniche.OthercompaniessuchasDellandFujitsuobtainedadvantagesofdistributionoroperationalefﬁciencythatenabledthemtoﬂourishincommoditymarkets(Kraemeretal.,2000).

Butsuchwasnotinthe“organizationalDNA”ofthesethreefocalcompanies.IBMhadledthecomputerindustryfor30years,supportingthehugeinfrastruc-tureofworld’slargestcomputercompany.Similarly,astheonlysurvivingPCstartupofthe8-bitera,Ap-plehadfromitsbirthpursuedago-it-alonestrategy.Finally,Sun,whilebuildinguponthesuccessofUnix

Table4

Opensourceplatformmilestones,1995–2002DateMay1995October1995March1996May1996October1996December1996June1998July1998September1998November1998December1998

December1998February1999March1999May1999July1999August1999September1999October1999December1999February2000

Industry

TorvaldsoncoverofForbesMicrosoftlistsLinuxasathreatinSECﬁling

Microsoft“Halloween”memostirscontroversy

IBM,Compaq,OracleandNovellinvestinRedHat

AfterIPO,RedHatreachesUS$4billionmarketcap

Microsoftposts“LinuxMyths”page,attacksLinux

Apple

StartsworkingonMkLinuxforPowerMacintosh

ReleasesMkLinux,givesaway20,000+CDstoISVs

BuysNeXTtoadaptMach/BSD-basedOS

EndsworkonMkLinux

AnnouncesDarwinopensourceOSandApplePublicSourceLicenseReleasesDarwinsourcecodeonApplewebserver

IBM

AnnouncesuseofApacheacrossentireserverline

AnnouncesDB2willsupportLinux

GermanIBMengineersportLinuxtoS/390mainframe

AnnouncesopensourceforJikesJavacompiler

SupportsLinuxforworkstationsandPCservers

Certiﬁes4LinuxversionsforPCservers

ReleasesDB2forLinux

LaunchesdeveloperWorksopensourcesite

PostsitsLinuxforS/390forfreedownload

Sun

Pre-releaseJavapostedforfreeInternetdownload

LicensesJavatoMicrosoft

SuesMicrosoftover“polluted”JavaAnnounces“100%PureJava”initiative

AbandonsISO/IECJTC1Javastandardseffort

IntroducesSunCommunitySourceLicense,planstoreleaseJavasourcecode

AnnouncessupportforLinuxonUltraSparcCPUs

Java2sourcecodepostedunderSCSL

BuysStarOfﬁce,revealsplanstoreleasesource

ReleasesSPARCchipdesignunderSCSL

AnnouncesitwillreleaseSolarisunderSCSL

AbandonsECMAstandardsprocessonJava;ResellsRedHatLinux

J.West/ResearchPolicy32(2003)1259–12851267Table4(Continued)DateIndustry

Apple

IBM

Sun

August2000OpenSourceDevelopmentLabReleasestoolsunderIBMPublicReplacesownGUIwithopenfunded

License

sourceGNOME

September2000ReleasespublicbetaofPurchasesCobalt,makerofLinuxDarwin-basedMacOSXserverappliances

October2000PostsStarOfﬁcesourceunderLesserGPLlicense

November2000RedHatdropsLinuxsupportforSunsystems

December2000AnnouncesIBMwillspendUS$1ReleasesSolarissourceunderSCSLbilliononLinuxin2001

license

January2001SettlesJavalawsuitwithMicrosoft

March2001Microsoftlaunches“SharedSource”ShipsMacOSX,basedonDarwinLaunches“Peace,LoveandLinux”initiative

andFreeBSD

adcampaign

June2001MicrosoftCEOBallmersaysHiresFreeBSDco-founder

“Linuxisacancer”

November2001LaunchesEclipseopensourceconsortium

April2002

HPwinsLinux-based

Withnon-proﬁtInternetSoftwaresupercomputercontractwithUSConsortium,co-sponsorsDepartmentofEnergy

OpenDarwin.org

May2002RetailsStarOfﬁceforUS$76

September2002

RedHatLinuxexpandssupporttoIntroducesLXseriesLinux-basedallIBMservers

PCservers

1268J.West/ResearchPolicy32(2003)1259–1285J.West/ResearchPolicy32(2003)1259–12851269

andtheopensystemsmovement,hadsoughttodiffer-entiateitselfbyhavingthemostcompleteandfullyfeaturedUnix-basedoperatingsystem.

UntiltheriseoftheIBMPC“clones”,IBMhadlargelycontrolleditsowndestinythroughpropri-etaryarchitectures.However,towinalliesfortheirdefactostandards,IBM—andtoalesser,degreeSunandApple—hadtriedtoleadformalanddefactostandardsorganizationstosupporttheirrespectivetechnologyinitiatives,aswithIBM’sDOS/VPCstandardinJapan(WestandDedrick,2000).

Suchalliancesweremoreproblematicforopensourcetechnologies.Inthedecentralized“bazaar”modelepitomizedbyLinux,therewasnocentraladministrativeauthoritywithwhichtonegotiate;tosomedegree,publisherssuchasRedHatandSuSEendedupﬁllingthisrole.5ThemorecentralizedcooperativeeffortssuchasFreeBSDorApache—the“cathedral”inthefamoustypologyofRaymond(1999)—providedamoreidentiﬁableauthorityfornegotiatingalliances.

4.2.Leveragingopennesswhilekeepingdifferentiation

TheopensourcemovementwasinpartframedasareactiontoMicrosoftanditsproprietarycontrolofthecomputerindustry,justastheopensystemsmovementadecadeearlierhadbeenareactiontotheproprietarycontrolofIBM.Assuch,themove-mentfoundnaturalafﬁnitywithMicrosoft’sthreemajorplatformcompetitors,aswellashardwaremakers—particularlyIntel—whobothbeneﬁtedfromMicrosoft’ssuccessbutalsowantedtoincreasetheirindependencefromit.Thusin2001,HP,IBM,IntelandNEClaunchedajointopensourceresearchlabinOregonandJapan;meanwhile,LinuxpublisherRedHathadalreadywonequityinvestmentsbyIn-tel,followedbyCompaq,Dell,HP,IBMandNovell(MacCormackandHerman,1999;WestandDedrick,2001).

TheproblemforIBM,AppleandSunwasthatbymakingsourcecodefreelyavailableandmodiﬁable,opensourceinherentlyreducedbarrierstoentryby

Inanalyzingthe“bazaar”archetypefortheLinuxproject,onemustrecognizetheongoingdefactoleadershipandcontrolexertedbyfounderLinusTorvalds.

rivalsandswitchingcostsbycustomers.Sodespitetheappealinglogicofmutualadversaries(“theenemyofmyenemyismyfriend”),apureopensourcestrategywouldeliminateeachcompany’shistoricsourceofdifferentiation,theirproprietarysoftware.Eachoftheﬁrmsfacedadilemmaofhowtoadaptanopensourcestrategysuitablefortheirrespectivecorecompetenciesandresources.

5.Apple:re-useandleverage

AppleComputer’sheydayhadcomewithitscre-ationofproprietarycomputerplatforms,but—likeIBM—itsonetimesupplierMicrosofthadbecomeaformidablerivalthatthreatenedthistraditionalsourceofadvantage.ThisforcedAppletoconsider(andembrace)somethingpreviouslyunthinkable:sharingtechnologythroughtheuseofopensource.

Apple’sadoptionofopensourcecameafterseveralfailedattemptstodevelopanewPCoperatingsystem,andseemedtoofferAppleawayoutofitstechno-logicaldead-end.In1995,ApplewastheﬁrstmajorcomputermakertosponsoraLinuximplementationforitsownhardware.TheexperimentalLinuxprojectwaseventuallydropped,butitwasreplacedbyanewOSthatcombinedauniquemixofproprietaryandopensourcecomponents.

5.1.Strategicpositioninmid-1990s

AlthoughthemostsuccessfulUSmakerof8-bitper-sonalcomputers,Applehadseveralfalsestartsbeforereleasingapopular16-bitPCwiththeMacintosh.TheMacintoshdifferentiateditselfwitheaseofusepro-videdbyaproprietarygraphicaluserinterface(GUI),butforvariousreasonslaggedMS-DOSanditsWin-dowsinadoption.6Finally,cuttingpricestorespondtoitsMS-DOSrivals,Applegrewmarketshareinmajormarketsintheearly1990sandenjoyedrecordmar-ketshareandrevenuesthroughitsﬁscalyearendingSeptember1995.However,theAugust1995releaseofWindows95effectivelyeliminateditseaseofusedifferentiation.Boththeactualandpredictedshiftindemandhelpedfueladownwardspiralthatinthenext

ThediscussionofApple’sproprietaryplatformstrategiesisadaptedfromWest(2000).

1270J.West/ResearchPolicy32(2003)1259–1285

2yearsbroughtnearlyUS$2billioninlossesandforcedresignationoftwoAppleCEOs.7

Ongoingimprovementsinoperationalefﬁciencyfrom1996to1998,thereturnofSteveJobsasCEOinAugust1997andtheintroductionofthepopulariMac1yearlateratleasttemporarilyquelledpredictionsofthecompany’simmediatedemise.Butthecompanyfacedthesamedilemmaasatthebeginningofthedecade—upgradingitscoreoperatingsystemtoin-corporatemodernmulti-processingandmemorypro-tectionfeatureswithoutrenderingitssoftwarelibraryobsolete.WhileMicrosofthadreleasedWindowsNTasaneventualreplacementforWindows95,Applewasstillusingimprovedversionsofits1984archi-tecture,designedforamachinewith128KBRAMandtwoﬂoppydrives.Moreseriously,thecompany’sdeclineinbothproﬁtsandR&DstaffmadeiteventoughertorespondtotheMicrosoftchallengethanithadbeenin1990.

Asaquickﬁx,in1996Appleevaluatedvariousalternativestojump-startitsfutureoperatingsystemdesign,includinglicensingWindowsNTandSun’sSolarisvariantofUnix.Intheend,itbypassedthepurchaseofthepromisingbutincompleteBeOStoacquireNeXT,acompanythatJobshadfoundedin1985afterleavingApple.TheacquisitionbroughtnotonlyNeXT’soperatingsystemtoApple,butJobsandhisR&Dteam,whichquicklyassumedkeypositionswithinthecompany.5.2.ShiftingtoUnix

ThesearchforamodernoperatingsystemwouldeventuallybringboththeUnixoperatingsystemandopensourcetothecoreofApple’slongtermplatformstrategy.InbuyingNeXT,ApplecastitsfuturewiththeNextStepoperatingsystemthatwastheacquiree’sprimaryasset.NextStepwasaUnixvariantthatcom-binedtheMachoperatingsystem“kernel”withothercomponentsfromBSDUnix.TothisNeXThadadded

TheconventionalanalysisattributesApple’sproblemsinthelate1980sand1990stoitsfailuretolicenseitsoperatingsystemtorivalspriortotheriseofWindowsin1991.Assessingsuchposthocadviceiscomplicatedbythevarietyofotherstrategicandex-ecutionerrorsduringthisperiod,includingnearly$2billionspentonunsuccessfuleffortsbothtoupgradeitscoretechnologyanddiversifyintonewmarketsegmentssuchashandheldcomputersandset-topboxes(West,2003).

variousextensions,includingagraphicaluserinter-face,softwaredevelopmentandsystemadministrationtools.

PriortotheNeXTacquisition,ApplealreadyhadexperiencewithUnix,offeringtheAT&T-basedA/UXoperatingsystemfrom1988to1995.ButbythetimeitwasabandonedafterashifttoIBM’sRISC-basedprocessors,A/UXhadattractedonly15,000users—atatimewhenApplewasselling4millionunitsayear(Hess,1995;Dataquest,1996).8

ApplehadexperimentedwithboththeMachOSandopensourcepriortotheNeXTpurchase.In1995,tohelpitsoperatingsystemsdesigneffortsitfundedaresearchprojecttoadapttheMachkernelforApplehardware.BecauseoftherisingpopularityofLinux,AppleselectedaMach–LinuxcombinationinsteadofthestandardMach–BSDcombination.Thisoperat-ingsystem,namedMkLinux,wasreleasedinaseriesofdeveloperreleasesfromMay1996untilthesum-merof1998,whentheprojectwasdestaffedbyAp-pleandhandledovertoitsbaseofuser-programmers.TheMachandLinuxsourcecodewerereleasedunderBSD-styleandGNU-styleopensourcelicenses.

InsteadofcontinuingwithLinux,the1997mergerwithNeXTcommittedAppletoadaptingitsexistingMacOStousetheBSD-basedNextSteptechnology.In1997itannounced“Rhapsody”—aMacintoshversionoftheexistingNeXTOS—whichshippedas“MacOSXServer”in1999.However,thecompany’smainfo-cuswasacomprehensiveupgradeoftheNeXTtech-nology,incorporatinganewGUIandanemulatorforolderMacOSprograms.This“MacOSX”wasorigi-nallypromisedforlate1999,wasreleasedinbetaforminSeptember2000andeventuallyshippedtousersinMarch2001.During2000and2001,theMacOSXuserandServercodebasesweremergedsothatbothusedthesamecoretechnologyanduserinterface.5.3.BuildinganewOSonopensourcepartsTheoriginalNextStepwasbasedonbothAT&TandBSDlicensedUnixcode,butby1999ApplereplacedthatcodewithFreeBSD,oneofthethreeopensourceIntel-basedversionsofUnix.Moresig-niﬁcantly,Appleannouncedthatitwasreleasingthe

From1996to1997,ApplesoldIBM’sAIXversionofUnix,inbetweencancelingA/UXandbuyingNeXT.

J.West/ResearchPolicy32(2003)1259–12851271

Mach,FreeBSDandsomeNeXTcomponentsasanewopensourceoperatingsystem,“Darwin”,whichout-sideexpertsdescribedasanewmemberoftheBSDfamilytree.

Darwin,ineffect,wasthecentralcoreofApple’sMacOSXServerandsubsequentMacOSX(Fig.1).Itprovidedprovenmulti-usermemorymanagementandprocesscontrolserviceswhichtheoriginalMacOShadlacked.Atthetime,Appleproclaimeditwas“leadingtheindustrybybecomingtheﬁrstmajorOSprovidertomakeit’s[sic]coreoperatingsystemavail-abletoopensourcedevelopers”(AppleComputer,1999).

ProjectmanagerErnestPrabhakarnotedthatacat-alystfortheDarwinstrategywaspressurefromlargeuniversitycustomerswithspecializednetworkingneeds:

Werealizedthatthepiecesthey’remostinterestedinarethemostcommoditized.Therewasn’tanyproprietarytechnologyaddedthatwehadtoworryaboutthemcopying....Westartedmakingthecase[that]weshouldjustopenthesourcecodeandre-leaseitasacompleteBSD-styleoperatingsystem.(Wayner,2000,p.175)

PrabhakarlatersaidthatApplesoughtto“embraceandenhance”existingopensourcetechnologies,butinsomecaseswould“embraceandlayer”bybuildingApple’sproprietarycodeontopofthepubliclysharedopensourcecode.

Themovewasproclaimedbyoptimistsaslegit-imizingopensourcedevelopment,inthataoncepro-prietaryplatformﬁrmwaswillingtoshareaportionofitscoreoperatingsystemwithoutsidedevelopers.However,otheropensourceadvocatesattackedsomeoftheexclusionsinApple’slicense,andthecontro-versydividedpreviouslyalliedleadersoftheopensourcemovement.Whilethemajorlicenseconcernswereresolvedwithinamonth,thedisputehurtopensourceadvocateswithinApplebydemonstratingthelackofasinglespokespersonwithintheopensourcecommunitywhocouldspeakonbehalfoftheentirecommunity(Shankland,1999;Wayner,2000,pp.162–163).

SomealsocomplainedaboutApple’schoiceoftheBSD-styleratherthanGPL-stylelicense,whichallowsacommercialﬁrmtotakethepublicsource,makepro-prietarymodiﬁcations,andtoreleasethatpartpub-lic,partprivateproductasaproprietarysolution(e.g.Leibovitch,2001).Suchisboththeinherentadvantage(forcommercialﬁrms)anddisadvantage(foropensourcepurists)oftheBSD-stylelicense.

ButthemostenduringcontroversycontinuedoverApple’sdecisiontoholdsomelayersofitsoperatingsystementirelyproprietary.MostofthepublicDarwinsourcewasaderivativeworkofthepublicFreeBSDandMach,andsoasapracticalmatterDarwinofferedthesamefunctionalityasitsFreeBSDcousin.Applehadheldoutthelargest(roughly75%)andmostvalu-ablepartsofMacOSX—itsgraphicaluserinterface,theNeXTandMacOSapplicationsupport—meaningthatDarwinwasnotacompleteGUIoperatingsys-temandthusoflittleinteresttoaverageusers.OtherApple-controlledtechnologies—suchasitsTrueTypefontsandQuickTimemultimediasoftware—weresimilarlyexcluded,preventingDarwin(andLinux)

Fig.1.ArchitecturallayersforApple’sMacOSXplatform.Source:AppleComputer(2000).

1272J.West/ResearchPolicy32(2003)1259–1285

usersfromusingthesetechnologieswithoutApplehardware.

InMarch2002,Applehelpedlaunchaneworga-nization,OpenDarwin.org.OpenDarwinmaintainedastand-aloneversionofDarwinoutsideApple’sdirectcontrol,withitsowndiscussiongroupsandbuglists,butsharingsourcecodewithApple.ByDecember2002ithad48identiﬁableuser-contributors.

Onpaper,Appleenjoyedthebestofallpossibleworlds.TheopensourceDarwinallowedittolever-ageoffthelargerBSDcommunitiestoincorporateen-hancementsinnetworkingandothertechnologies,andtoportUnix-basedapplicationssuchaswebandmailservers.Darwinalsoprovidedlow-leveldocumenta-tiontothird-partyhardwarevendors,freeingAppletoconcentrateitssupporteffortsonapplicationsoft-ware.Applealsoretaineddifferentiationinthetradi-tionalareaswhereithadmatteredmost—ingraphicsandeaseofuseforitscoremarketsingraphicaldesignandeducation.

However,byopeningonlypartofitstechnology—largelycorrespondingtotheexistingFreeBSD—Applemadeitlessvaluabletouser-contributors.ThefewerusersthatcontributedtotheDarwinsources,thelessbeneﬁtApplerealizedfromitsopensourcestrategy.

6.IBM:fromplatformstoapplications

Despiteitswellpublicizedtravailsofthe1980sand1990s,IBMremainstheworld’slargestcomputercompany,asmeasuredbothbytotalsalesandemploy-ees.ItwasalsothemostaggressiveofanyincumbentcomputermakerinembracingLinuxandopensource.AsheannouncedIBM’s2001planstoinvestUS$1billioninLinux,thethenCEO,LouisGerstner,pre-dictedanendtotheeraofproprietaryplatformsthathiscompanyhadspawned:

Themovementtostandards-basedcomputingissoinexorable,IbelieveSun—andEMCandMicrosoftforthatmatter—isrunningthelastbigproprietaryplaywe’llseeinthisindustryforagoodlongwhile.(Wilcox,2000)

WhileGerstnermightsecretlypreferareturntoIBM-controlledproprietaryindustry,therealitywasthathefoundthatanopensourceworld—wherehard-

warevendorsandcustomersallhadfullcontrolovercrucialsystemsoftware—waspreferableaproprietaryindustrycontrolledbyMicrosoft.6.1.Strategicpositioninmid-1990s

IBMwasbestknownforcreatingtheﬁrstcomputerplatformwithitsS/360proprietarymainframes,anditssubsequentAS/400minicomputerplatform.Manyargued,however,thatitsmostsuccessfularchitecturewastheonethatgotaway:theIBMPC.TheshiftofleadershipwastraumaticforIBM.Intheearly1990s,IBMspentbillionsofdollarsonOS/2andjointven-tureswithvariousindustryrivals—allinanunsuccess-fulattempttore-assertitsproprietaryleadershipofthecomputerindustry,oratleastbreakfreeofdepen-denceonMicrosoft.By1996,thehatredofthe“EvilEmpire”withinIBMborderedontheprofane(Garr,1999,pp.187–188).

AfterlosingthePCOSwar,IBM’sleadershipwasconﬁnedtothosesegments(largecomputersystems)thatwereenduringasystemicdecline.InmorerapidlygrowingPCandworkstationmarkets,itwasanalsoran.Inresponse,thecompanysoughttorepositionthedemandformainframecomputers,ﬁrstbypromotingclient/serverarchitectureandlaterbydevelopingsoft-waretoenablethemtoserveasmassivewebservers.Meanwhile,itplacedincreasingemphasisonthesaleofsoftwareandservices,winningbusinessbasedonitsunmatchedabilitytoofferacompleteend-to-end“turnkey”solution.

6.2.PhaseI:applicationsoftware

IBM’sﬁrstmajoropensourceinitiativecameinitseffortstointegratecorporatemainframes(withtheirvastlegacydatabases)todirectlysupporte-commerceandintranetinitiatives.InJune1998,itunveileditsWebSphereproductfamily,whichbuiltupontheApacheopensourcewebpageserver.IBMhadbe-gunwithitsowninternallydevelopedwebserver,butadoptedtheInternet’smostpopularwebserverafterfailureofnegotiationswithNetscapeoverlicensingitsproprietarywebserver(McKay,1998).

IBM’seffortstoadaptApachetomeetitsspeciﬁcneedssetapatternforcollaborationinitssponsorshipofsubsequentopensourceefforts.IBMhelpedfundApache’sadaptationforuseonWindowsNT,because

J.West/ResearchPolicy32(2003)1259–12851273

itwascentraltoitsWebSpherestrategy(MoltzenandBurke,1998).Meanwhile,IBMengineerscontributedcodebackforusebyallApacheimplementations,andIBMhiredoneofthekeyopensourcedeveloperstoactasapermanentliaison(Wayner,2000,pp.181–183).IBMfoundthatworkingwith(largelyuserdriven)opensourcegroupprovidedmoreﬂexibilitythanusingsomeoneelse’sproprietarysolution,alessonitwouldlaterapplytoLinux.

WebSpherewasalsotheﬁrstIBMapplicationtobemadeavailableonvariousLinuxplatforms.IBMaddedaLinuxversionofitsDB2database,whichinDecember1998wasreleasedinbetaformonIBM’swebsite;IBMlaterportedathirdmajorproduct,itsLotusDominogroupware.AswithotherUnixappli-cationvendors,IBMfoundthatofferingaLinuxver-sionrequiredacomparativelysmallinvestment.

TheWebSphereproductindirectlyledtoIBM’sNovember2001formationofEclipse,anindependentopensourceconsortiumtodevelopcommonsoftwaretools.IBMdonatedsourcecodeitvaluedatUS$40milliontolaunchtheconsortium,whicheventuallygrewtoincludeFujitsu,HP,Oracle,RedHatandSAP.Sun(withitscompetingJavaBeans)wasnotablyab-sent.

WhileIBM’sinitialfocuswasondevelopingweb-enabledapplicationsinJavaandHTML,theEclipseprojectbraggedthatitwasa“universaltoolplatform...foranythingandnothinginparticular”(Eclipse.org,2002).TheIBM-developedarchitec-tureallowedthird-partydeveloperstowriteplug-inmodulestosupportvariousprogramminglanguages(C,C++,Cobol),ﬁleformatsandexternalsoftwareproducts(likeIBM’sWebSphereApplicationServer).ThecodewaslicensedundertheCommonPublicLicense,anopensourcelicensewhichlikeBSD—andunlikeGPL—allowedcommercialdistributionofderivativeworks.AsEclipseevolved,IBMmergedupdatesfromitintoitscommercialproduct,Web-SphereStudioWorkbench.InDecember2002,IBMannouncedaUS$2.1billionpurchaseofRationalSoftware,anEclipseconsortiummemberthatsoldap-plicationdevelopmenttools.

Overall,IBM’sdevelopmentofapplicationsusingopensourcesoftwarehadthreecommonthreads.First,IBMacceptedcommodizationofcertainlayersofitsapplicationarchitectureandwasthuswaswillingtocollaboratewithopensourcesoftwareprogrammersto

makeasharedtechnologyavailabletoall;theselayerstypicallyimplementedopenInternetstandardswhichofferedlessopportunitiesfordifferentiation.Second,inmanycasesthesharedsoftwarecompetedwithpro-prietarysolutionsdevelopedbyMicrosoftusingitsUS$4billionannualR&Dbudget,suchasitsInternetInformationServiceswebserver.Finally,thesharedsoftwarewasreleasedunderanon-GPLlicenseal-lowingIBMtoretaintechnologyormakeproprietaryenhancements.

6.3.PhaseII:systemsoftware

Unlikewithapplications,IBM’sdirectsupportforLinuxasareplacementoperatingsystemwaslongincoming.Throughout1998,IBM’scorporatestrate-gistshadnointentionofprovidingorsupportingLinux,asitwouldreducedifferentiationandalsothreatenitshigh-marginproprietaryoperatingsys-temsales.SoinsteadofbeingofﬁciallysponsoredbydevelopersofIBM’sS/390mainframe,Linuxwassuccessfully“ported”inlate1998throughun-sanctionedeffortofprogrammersatIBMGermany(Hall,2001).ThisinternalversionofS/390LinuxwaseventuallyreleasedforcustomerdownloadsinFebruary2000.Subsequently,thechangestostan-dardLinuxtosupporttheS/390weremadeavailablebothonIBM’swebsiteandthroughcommercialdistributorsRedHat,SuSEandTurboLinux.IBMalsoendorsedexistingLinuxdistributionsforitsPCservers.

Inearly2000,IBMannounceditwouldsupportLinuxacrossitsentirerangeofservers,fromPCsuptothelargestmainframe(Table5).Atthesametime,areorganizationeliminateditsInternetbusinessunit,andreassigneditshead,IrvingWladawsky–Berger,toheadacombinedUnix–Linuxunit.ThatgroupwaslocatedwithinIBM’sEnterpriseSystemsdivision—clearlytargetingIBM’straditionallargecorporatecustomersratherthanthePC-basedInternetserviceprovidersthatuptothatpointhadbeenthelargestmarketforLinux.

Beyondthelevelofcustomerdemand(or“marketpull”),endorsingLinuxgaveIBMkeystrategicad-vantages.First,LinuxprovidedacommonsetofAPIsacrossitsentireproductline,providingauniﬁedar-chitectureforsoftwaredevelopers.Second,thecom-parativelyimmature(yetcomplex)operatingsystem

1274J.West/ResearchPolicy32(2003)1259–1285

Table5

IBM’ssupportedplatformsbeforeandafterLinuxadoptionEarlierplatformsCategory

MainframeMinicomputerWorkstationServerPC

Desktop/laptopPCPDA

Revisedplatforms

Hardware

IBMS/390AS/400RS/6000Netﬁnity

NetVista,ThinkPadWorkPad

Operatingsystem

OS/390OS/400AIX

WindowsNT,Windows2000Windows95/98/MEPalmOS

RebrandingzSeriesiSeriespSeriesxSeriesNANA

OperatingsystemaLinuxLinuxLinuxLinuxNANA

TraditionalOSalsoavailableonrebrandedplatforms.

requiredsupportservices,atraditionalIBMstrength,asWladawsky–Bergerlaterexplained:

We’veweddedourselvestotheintegrationofthesolution,thenotionbeingthattheInternetande-businesssolutionsaremoreimportantthananyparticularcomponent.Andasaresult,we’vechangedallourbusinessmodelssothattheinte-grationofthepieceshasbecomemoreimportantthananyonepiece.(Cooper,2001)

Finally,theopensourceoperatingsystemallowedIBMtomakechangestoimproveitshardwarediffer-entiationforenterprisecustomers.Aschieftechnol-ogyofﬁcerforPCserversnoted,Linux“hasgivenIBManopportunitywedidnothavebeforetoplaytoourstrengths,whichisavailabilityandreliability”(Shankland,2002).

Wladawsky–Bergeralsosaidthattheopensourcemodelwasalogicalextensionofthelong-standingIBMresearchculture.Towardsthatend,inAugust2000IBM’sspecialdeveloperWorksopensourcewebsitereleasedthesourcecodefromtwolargeIBMre-searchprojects,theAndrewﬁlesystem(latercalledOpenAFS)anditsJikesJavacompiler.Bothwerere-leasedundertheIBMPublicLicense,anantecedentoftheCommonPublicLicense.7.Sun:openingnewplatforms

Sunwasfoundedin1982tomakeengineeringworkstations,andbythelate1980shadoutlasteditsrivalApollotoleadthemarket.WhileitsSunOS(laterSolaris)platformwasbasedonthesameUNIXoperatingsystemadoptedbymostofitsworksta-tionrivals,Sunsuccessfullydifferentiateditselffrom

otherrivalsthroughongoingenhancementsinitsop-eratingsystem,particularlywithitssupportfordatanetworking(GarudandKumaraswamy,1993).

Asthe1990sended,Sunfacedthreatstoitscorebusiness,byopensourceLinuxontheonehandandMicrosoft’sWindowsNTontheother.Atthesametime,itsoughttoretainitstraditionalcontroloftheoperatingsystemandothertechnologythathadfueleditssuccessintheworkstationandservermarket.7.1.Strategicpositioninmid-1990s

Inthemid-1990s,Sunheldastrongpositioninworkstations.Meanwhile,itwaswellpositionedtocapitalizeonrapidlygrowingindustrydemandformidrangeserversandthosethatrequirednetworkingandInternetsupport.WhileSun’smarketinghadhis-toricallyemphasizedsupportfor“openarchitecture”,itusedproprietaryextensionstoUnixsoftwaretodif-ferentiateitselffromworkstationrivalssuchasHP,IBMandDEC(laterCompaq).9

Toimproveadoption,itlicenseditsworkstationandOStechnologytocustomersandcomplemen-tors;thisincludedasmallnumberofmakersof“clone”products,mostnotablyinJapan(GarudandKumaraswamy,1993).Atthesametime,Sunre-tainedfullcontrolofthearchitecture,allowingittorapidlyevolvethetechnologyratherthannegotiatewithstandardscommittees.Assuch,Sun’sstrategymorecloselyﬁtthe“proprietarybutopen”modelofMorrisandFerguson(1993)thandidMicrosoft.

However,thedifferencesinUnixAPIsincreasedcostsforuserandsoftwaredevelopers,eventuallyforcingvendorstoagreeonacommonGUIandsystemspeciﬁcationsinthemid-1990s,endingthe“Unixwars”.

J.West/ResearchPolicy32(2003)1259–12851275

ByconcentratingonUnix-basedsystemsandig-noringPCs,inthemid-1990sSunheldauniquepositioninthecomputerindustrywithrespecttoMicrosoft.MostcomputercompanieslicensedoneofMicrosoft’sWindowsoperatingsystemsforservers,PCs,laptopsorhandheldcomputers;whileApplenotablydidnot,itbundledMicrosoft’swebbrowserandactivelycourteditsOfﬁceapplicationsuite.Assuch,nomatterhowmuchsomeinitiativesofIBMorApple(orHPorCompaq)mightconﬂictwithMicrosoft,atothertimestheywereMicrosoftal-lies.EvenbitterrivalsAmericaOnlineandNet-scapemadeWindowssupporttheirtopapplicationspriority.

Bycontrast,Sunlackedsucha“co-opetition”re-lationshipwithMicrosoft.Adecadeearlier,thetwoﬁrmshadlittleoverlapwithcompletelydifferenttech-nologiesandcustomers.However,industrytrends—particularlytheshiftoftheInternetfromaresearchnetworktoaconsumerone—hadledthemtobecomedirectcompetitors(Goff,1999).TheWindowsNTserveroperatingsystemwasdirectlyaimedattheUnixserverbusinessledbySun,whileanysuccessofMi-crosoftinestablishingproprietaryInternetprotocolswouldcomeatSun’sexpense.Meanwhile,SundidnotshipMicrosoft’soperatingsystemorapplications.Thus,itwasnotsurprisingthatSun’sco-founderandlong-timeCEOScottMcNealywasoneofMicrosoft’sharshestandmostvocalcritics,bothbecauseofthetwocompany’sconﬂictinggoalsandbecauseofthelackofdependenceonMicrosoftforanykeytechnology.

BothWindowsNT(laterWindows2000andXP)andLinuxposedalow-costthreattoSun’slucrativeserverbusiness:bothwerebasedonhigh-volumeIntelprocessors,whoseperformancewasincreasingmorerapidlythanthatofSun’sproprietaryRISCproces-sors.Inanotherdimension,WindowsandLinuxwereattackingSunfromoppositesides:theformerrepre-sentedamoreproprietaryapproachundercontrolofastrong,centralizedrival,whilethelatterofferedgreateropennessthatwassupported(initially)byadiffusegroupofhobbyists.

7.2.Strategy1:newplatforms

Sun’sprimarystrategyduringthelate1990swastoestablishnewplatformsindependentofMicrosoft

thatwouldlimit(ifnotreduce)Microsoft’scontrolofindustrystandards.

MostofSun’seffortswenttowardsestablishingJavaasanewplatformwithacommonsetofAPIsavailableonawiderangeofcomputersystems,undertheslogan“writeonce,runanywhere”.AnearlyprototypeofJavawasmadeavailableforuserdownloadsinMay1995,andlicensedbymostmajorcomputercompaniesoverthenextyear(Garudetal.,2002).Sunmounteda4-yearlawsuitaccusingMicrosoftoftryingtohinderJava’ssuccess,alawsuitsettledoutofcourtinJanuary2001.

Afterdistributingpreviewsofitstechnologyandgeneratinggreatinterest,Sunspent3years(1997–2000)tryingtogetJavaestablishedasadejurestandardalbeitunderSun’scontrol,ﬁrstattheISO/IECJointTechnical10Committee1(JTC1)andthenlateratECMA.Botheffortswerewithdrawnaftervigorousandwell-ﬁnancedoppositionfromcompetingcomputermakers,particularlyMicrosoftandHP(Egyedi,2001).Sunalsoobjectedtothepro-visionsthatwouldhaverequiredittosurrenderIPRtothestandardizationcommittee.

Inasecondmajorinitiative,inAugust1999SunspentUS$73.5milliontoacquiretheGermanmakerofStarOfﬁce,acloneofMicrosoftOfﬁce.AsSunlaternoted“ItiscriticalforallofSun’scustomersthattherebeopen,viable,cutting-edgeofﬁceproductivitysoftwareavailabletorunintheheterogeneousnetworkandacrossallplatforms”(SunMicrosystems,2001).Thecompanyalsostatedadesiretoshifttheindustryfromthepurchaseofsoftwareproductstotherentalofnetwork-intensiveapplicationservices(whichwouldrequiremoreSunsystemstoimplement).

Toalargedegree,Sun’spreoccupationwithitsof-fensivestrategyagainstMicrosoftcontributedtoitsfailuretodefendagainstLinux.Forexample,ina3,000+word1996interview,Sun’sCEOdidnotmen-tionLinuxoropensourcecodeatall.Instead,hereiteratedSun’shistoricpushforUnix-basedopensystemsoverproprietaryplatforms:

Q:NowMicrosoft’srecentlicensingofJavaseemslikeasolidvoteofconﬁdenceinthetechnology,but

ECMAwasfoundedin1961astheEuropeanComputerMan-ufacturersAssociation,butin1994switchedtoitsacronyminhopesofincreasingitsglobalinﬂuence.

1276J.West/ResearchPolicy32(2003)1259–1285

areyouworriedMicrosoftmighttrytopositionJavaasjustanotherlanguage,orburyitunderActiveX?McNealy:We’realwaysworriedpeoplewilltryandhijackthestandardsonthenetwork...TherealbeautyoftheNetisallinterfacesareopen,they’remulti-vendorandyoucanpublishyourdataorpub-lishyourapplicationonceandknowitwillrunoneverything.Thatmakesitalotmorecompetitiveandlowersthepricetotheuser.That’snotexactlywhatallthetechnologycompanieswanttodo.Theywanttogetyoulockedin.(Taylor,1996)Ironically,theeffortsofMicrosoftandotherpropri-etarycomputermakerstoderailSun’splatformstrate-giesledittowardsitsﬁrstopensourcestrategies,whichimproveditsabilitybothtocompeteandcoop-eratewithopensourcesoftware.7.3.Strategy2:partly-opensource

InsteadofrespondingtoLinux,Sun’sﬁrstopensourcestrategiesfocusedonitscompetitionwithMicrosoftingettingcoretechnologiesadoptedbyusersandsoftwaredevelopers.TheevolutionofthesestrategiesbeganwithJava,extendedtoStarOfﬁceandeventuallyreacheditscoreSolarisoperatingsystem.Inadditiontoitspro-activestrategies,Sunalsore-actedtopressuretoincreasetheaccesstoitsJavacodeexertedbylicenseesandstandardizationcommittees.TheefforttodeﬁneSun’sopensourcestrategywasledbychiefscientistBillJoy,whohadbeenoneoftheleadingengineersintheBerkeleyUnixgroupbeforeleavingtoco-foundSun(Kim,1999).Themostradi-caloption—whichwouldpleasethelargestnumberofopensourceadvocates—wouldhavebeentousetheGNUPublicLicense,inwhichallchangesmadebySun,itscustomersorcompetitorsmustbesharedwitheveryone.Whilethiswouldfueladoption,itposedrealconcernsaboutappropriatingthereturnsofSun’sR&Dinvestment,asJoyexplained:

Ican’tlicenseallofSun’sintellectualpropertyun-dertheGPL,becauseitjustwon’twork.Idon’tseeanyreasonwhyIshouldgivesomebodywho’sdo-ingcommercialreuseunfetteredaccesstostuffthatcostmemillionsofdollarstodo.We’respendingoverabilliondollarsayearinresearch.Ican’tjustthrowitallonthestreet...

IfImakecodeavailableundertheGPL,I’lllosecontrolofit....TheGPLjustdoesn’tsolvemybusinessproblematSun.Iwouldlikeallofourintellectualpropertytobeavailableinsourceform,butIcan’teconomicallydothatundertheGPL.(Kim,1999)

Instead,inFebruary1999SunreleasedJavasourcecodeunderwhatitcalledtheSunCommunitySourceLicense,ahybridbetweenatraditionalproprietaryli-censeandaBSD-styleopensourcelicense.Theli-censehadfourbasicelements:(1)righttomodifythesourcecode;(2)royaltyfreedistributioninopensourceprojects;(3)royaltiesforcommercialredistri-bution;and(4)testingrequirementtomaintaincom-patibilityandpreventforking(Loukides,1999).FromSun’sstandpoint,thelicense:“providesprotectionforintellectualproperty,...guaranteesstructuredinno-vationwithinasingleresponsibleorganization,[andprovides]clearcontrolovercompatibility”(GabrielandJoy,1998).TheSunlicenseisinnovativeinitsgovernance:theSCSLprocessesinmanywaysresem-bleaformalstandardsconsortiumordejurestandard-izationcommitteemorethanthe“bazaar”associatedwithdecentralizedLinuxdevelopment.

IfSunhadconcernoverrivalsandcontrolwithJava,thesewereevengreaterwithSolaris,itsUnix-basedoperatingsystemthathadprovideditskeydifferenti-ationintheworkstationandservermarket.However,inOctober1999itannouncedthatitwouldreleaseSolarisundertherestrictionsoftheSCSL,andinDe-cember2000ﬁnallydidso.

SunadoptedadifferentstrategyforStarOfﬁce;inOctober2000itreleasedallsourcecodetoanewor-ganization,OpenOfﬁce.org.Thecodewaslicensedinawaythatguaranteedthesourcewouldalwaysremainpublic,butalloweditsuseincommercialproductsbySunoranyoneelse.11Afternearlyadecadeofdevel-opment,StarOfﬁcebadlylaggedMicrosoft’sproductinfeatures,compatibilityandreliability.The“bazaar”communityofopensourceuser-developerswaside-allysuitedforaddressingsuchconcerns,asSunex-plained:

Technically,theLesserGNUPublicLicensehassomecharac-teristicsincommonwiththeGPL,butinpracticaltermsitwasmoresimilartoanApache-stylelicenseinthatitlackedthe“viral”provisionsoftheGPL.

J.West/ResearchPolicy32(2003)1259–12851277

Byengagingtheenergyandcreativityofdevelop-ersworldwide,wewillacceleratetheadditionofinnovativefeaturesandimprovedintegrationwithotherproducts.MakingthesourcecodeavailablealsoenablestheStarOfﬁcesoftwarefunctionalitytobeportedtoawiderrangeofsystems.(SunMicrosystems,2001)

In2002,SunreleasedaUS$76commercialversionofStarOfﬁce,sharingcodewithOpenOfﬁce,whichremainedanongoingopensourceproject.BothSunandtheopensourcecommunityretainedacommongoal,beingabletoaccessbusinessdocumentscom-patiblewithMicrosoftOfﬁceandthedominantWin-dowsstandard.

7.4.Strategy3:ifyoucan’tbeatthem,jointhemComparedtoitsotherworkstationrivals—particularlyHPandIBM—SundidlittletoembraceLinux.From1998to2000itprovidedtechnicalassistancetooutsidegroupsportingLinuxtorunonits32-bitSPARCand-bitUltraSparclineofSolaris-basedsystems.ItdidnotsellsystemswithLinuxpre-installed,leavingthattovalue-addedre-sellers.InfactthedemandforLinuxonSunhardwarewassoweakthatinNovember2000aleadingLinuxdistributor,RedHat,canceled12

futuredevelopmentofLinuxforSuncomputers.SunwasinitiallyambivalentaboutthesuccessofLinux.Ontheonehand,increasingadoptionofUnix-basedsystemsreducedMicrosoft’sinﬂuenceandimprovedavailabilityofUnix-relatedsoftware,trainingandengineers.Ontheotherhand,SunhadchargedapremiumtoitsworkstationrivalsbasedonitssuperiorSolarissoftware,soaworldwhereLinuxwasthenormwouldeliminatethatadvantage(Rosenberg,2000).Asoneanalystputit:“SunisgivingLinuxsomerhetoric,butSundoesnotwantLinuxtotakeoff”(ScannellandGardner,1999).Eventually,inthefaceofcompetitioninthelow-endserverbusiness,SunadoptedLinuxasaserveroperatingsystem,albeittoalesserdegreethanIBM.In2000,SunpurchasedCobalt,amakeroflow-endLinuxserverappliances.Twoyearslater,

Itshouldbenoted,however,thatSun’sInternetserversincor-poratedthesameopensourceapplications(suchasApacheandsendmail)astheotherUnixandLinux-basedservers.

SunannouncedplansforitsownbrandedIntel-basedservers,runningLinuxorSolaris.

8.EffectofopensourceonplatformstrategiesThepopularityofopensourceoperatingsystemscreatedbothproblemsandopportunitiesforpropri-etaryplatformvendors.Disclosingsoftwaretechnol-ogythroughopensourcelicenseswouldnaturallyleadtothecommodizationofsuchsoftware.Notsurpris-ingly,theresponseofleadingindustryﬁrmsvariedde-pendingonwhethertheyhadusedsoftwareasasourceofcompetitiveadvantage,andwhethertheyretainedothersourcesofcompetitiveadvantage.

8.1.ComparingstrategiesbyApple,IBMandSunWhileIBMwasoneoftheworld’stopsoftwarevendors,itssoftwarewasnormallysoldaspartofacombinedsolutionwithitsownhardware.Itcontin-uedtodifferentiateitselfbasedonservicessuchasin-tegration,servicesthatwouldprovideswitchingcostsifitadoptedcommoditysoftware.Also,itskeyrev-enueswereinmainframeandmidrangesystemswheretherewerefewremainingcompetitors.Undersuchconditions—witheitherservicedifferentiationorfewviablecompetitors—softwarecommoditizationwouldbelessofaconcern,andanythingthatreducescostsorincreasesdemandisanunvarnishedplus.Infact,LinuxofferedIBMsomethingithadneverhad—acommonsetofsoftwareAPIsacrossitsentireproductline.LinuxsuccesseswouldhurtAppleandSun,whichhistoricallybundledaveragehardwarewithbetterthanaverageoperatingsystems.AppleandSunfaceddi-minishedproﬁtmarginsiftheysharedthesamesoft-wareastheirrivals(orviceversa).ButunlikeIBM,bothAppleandSunreleasedsourcecodefromtheirprimaryoperatingsystem.Thetwoﬁrmsadoptedtwodifferentapproaches:openingpartsversuspartlyopen.Inpartbybuildingoncodethatwasalreadyopensource,ApplechosetograntallrightstoasubsetofitsnewOSXoperatingsystem.Meanwhile,SunreleasedtheentiresourceofJavaandSolarisunderrestrictiveterms—theformertoimproveadoption,thelatterinresponsetocompetitionfromtheopensourceLinux.AllthreeﬁrmssoughttomaintaincontroloftheirproprietaryOSandothertechnologies,inparttoassure

1278J.West/ResearchPolicy32(2003)1259–1285

thattheywouldcontinuetoevolveandremaincom-petitive.Theyalsohadspeciﬁcconcernsaboutaidingrivalsandanhistoricaversiontosharingproﬁtswithothersintheirrespectivevaluechains.

Priortoembracingopensourcestrategies,allthreecompanieshadextendedthevalueoftheirplatformsthroughproprietaryapplicationsand“middleware”.Suchsoftwarehadenabledthemtoservetheirrespec-tivemarkets—largecorporateserversforIBMandSunandmultimedia-savvyconsumersforApple.Byre-tainingthissoftwareasproprietaryanduniquetotheirrespectiveplatforms,theﬁrmswereabletoretainatleastsomedifferentiationrelativetobothproprietaryandopensourcecompetitors.8.2.Microsoft’sresponse

Oftheindustry’slargestﬁrms,Microsoftclearlyhadthemosttolosebythehavingfreesoftwaresup-plantcommercialoperatingsystemsandapplicationsoftware;itlackedthehardwareandservicesrevenuestoreplacesoftwaresaleslosttofreesoftware.Ontheotherhand,Microsoft’sproprietaryplatformstrategiescontinuedtobesuccessfulandthusitfacedtheleastpressuretoadoptanopenstrategy.However,therisingpublicityassociatedwithopensourceandthepoten-tialshiftofservercustomersfromWindowstoLinuxforcedittorespond.

Intheﬁrsthalfof2001keyMicrosoftexecutivespubliclyattackedthemovement,particularlythe“viral”natureoftheGPL(WestandDedrick,2001);thecompanylaterclariﬁeditspositiontoemphasizesupportforBSD-stylelicenses.Microsoftalsoun-veileditsownformoflicensedsourcecodedisclosurecalled“SharedSource”.By2002,itsstrategyhadevolvedtoallowPCvendors,third-partydevelopersandlargeend-userstoviewbutnotmodifythesourcetoWindows.Towintheheartsandmindsofaca-demics,Microsoftalsoalloweduniversitiestobothviewandmodifythesourceforinternalresearch.TheroleofMicrosoftwasclearlyparamountintheopensourcelicensestrategiesadoptedbybothnon-proﬁtandcorporatesoftwaredevelopers.LinuxdevelopersandotherbackersoftheGNUlicenseof-tencitedMicrosoft’sdecisiontouseBSDnetworkingtechnologiesinWindows,thenenhancethosetech-nologiesinawaythatmadeWindowsincompatiblewithUnix-likesystems.AsSun’sBillJoycomplained:

ThetoppredatornowisMicrosoft.Wedidn’thaveatoppredatorbackwhenIdidTCP/IP.Whenyouhaveapersonwithunlimitedfundswhoisclearlyfocusedondestroyingthevaluepropositionofwhatyou’redoing,you’dbeafoolnottoaccountfortheminthestrategythatyouadopted.(Kim,1999)9.Discussion

Theuseofmultiplequalitativecasestudiespro-videsarichopportunityforbuildingtheoryinemer-gentareasthatisgroundedinempiricaldata.ThissectionusesinductiontogeneralizetheobservedopensourceplatformstrategiesofproprietaryﬁrmsintobroadertheoreticalpredictionsaboutcompetitivestrategiesforITplatforms,andsuggestareasforfutureresearch.

Suchtheoryalwaysrunstheriskofbeingidiosyn-craticandnotgeneralizabletotheentirepopulation(Eisenhardt,19).Inthiscase,bystudyingﬁrmsthatwerepreviouslysuccessfulwithproprietarystrate-gies,suchﬁndingsmaynotbeapplicabletoﬁrmsthatunsuccessfullypursuedproprietarystrategies,ordenovoentrantsthatlackpriorplatformcapabili-tiesuponwhichtheycanbuild.Thereisalsotheriskofattemptingtogeneralizefromastill-emergentprocess:theadoptionofopensource—bothbybusi-nessend-usersandproprietaryhardwarecompanies—isstillcomparativelyrecentphenomenon.Anyorallofthecompaniesstudiedcouldfailintheirefforts,orﬁndgreatersuccessbyreturningtomoreproprietarystrategies.

9.1.Shiftingfromproprietarytoopensourcestrategies

Thestudysuggestsathreestageevolutionofpro-prietaryplatformvendorstotheuseofopensource.9.1.1.Proprietaryplatforms

Aswithotherindustries,computerindustrypi-oneersbeganbyverticallyintegratingtodeliveracompleteproprietaryplatformsolution.Wheneverpossible,theﬁrmspreferproprietaryplatformstrate-gies,becausetheyprovidebetterbarrierstoimitationandbettermargins.Butasnotedearlier,thisstrategymayonlybeavailabletooneortwomarketleaders.

J.West/ResearchPolicy32(2003)1259–12851279

9.1.2.Openstandards

FormanyITvendors,theuseofproprietaryplat-formstrategybecomesinfeasibleforsomecombi-nationoftechnicalandeconomicreasons,andtheymodifytheirplatformstoincorporateopenstandardsthataresharedwithoneormorecompetitors.Amongthemotivations:13

•marketsharelowerthantheminimumefﬁcientscalenecessarytosupportproprietaryR&D;

•notenoughmarketpowertoresistbuyerdemandsforopenstandards;

•“tipping”ofthestandardscontestinfavoroftheopenstandard,makingitinfeasibletoestablish(ormaintain)aproprietarystandard;

•adecisiontoacceptcommodizationoftheparticulararchitecturallayerandshiftcompetitiveadvantagetoanotherarchitecturallayer.ThesecriteriaareconsistentwithvonBurg’s(2001)analysisoftheadoptionofEthernetasthepreferrednetworkingtechnologybybothproprietaryandopensystemsvendors.

Whileshiftingtosuchsharedplatformsmaybethemostcost-effectivesolution,itcanbedifﬁcultwhenitrunscontrarytothecorporatecultureandpreviouslyvaluedcorecompetencies.TheexperienceofApple,IBMandSunsuggeststhatshiftingtoevenapartly-openarchitecturemayrequireamajorexternalshocktoforceﬁrmstorelinquishpreviousinnovation-drivendifferentiationstrategies.

Theshiftalsocontainsunexpectedpitfalls.Firmswithasuccessfulproprietaryarchitectureareabletosimplifytheirtechnicalandbusinessdecisions,be-causetheycontroltheirenvironmentanddonothavetointeroperatewiththerestoftheworld.Whentheproprietarystrategyfails,ﬁrmsareforcedtoworkwithopenstandardstoachieveinteroperability,andsuchinteroperabilitybothrequiresadditionaltechni-caleffortsandalsoreducesthelock-inofexistingcustomers.Forexample,facedwithplummetingmar-ketshare,Appleabandoneditsproprietaryperipheralinterfacestandardsandswitchedtoindustrywidestan-dards,discontinuingmostoftheproprietaryperipher-13

InthecaseofthePCindustry,thesharedmultivendorarchi-tecturallayersareproprietarytechnologiesprovidedbyMicrosoftandIntel,whichaddressessomebutnotallofthepressuresforopenmultivendorstandards.

alsthatonceaccountedformuchofitsrevenues(West,2000).

HavinglostvariousplatformbattleswithMicrosoft,bothAppleandIBMhavebeenforcedmorethaneverbeforetoshiftfromproprietarytoopenstandards.Sun’sbusinessmodelhadalwaysrequiredco-existingwithopenstandards,butnowhasrelinquishedcon-trolofmoreplatformstandardsto“competeonimplementations”.

Withoutinnovationandproprietarylock-intopro-videbarrierstoentryandimitation,invariablyﬁrmswillﬁnditdifﬁculttoachievecompetitiveadvantagewiththesenewstrategies.Amongthefunctionalstrate-giesthatthethreeﬁrmsusedincludemarketing,cus-tomerservice,productdesign,engineeringefﬁciencyandleveragingpreviouslyestablishbrandnamerepu-tations;thelong-termviabilityofallthesestrategieshaveyettobeproven.

9.1.3.Opensources

Thetransitiontoanopensourceplatformstrategyisacontinuationofthattoopensystems,drivenbythemanyofthesamefactors.Opensource,however,eliminatestheabilityofvendorstocompetebasedonimplementationssincethedetailsofanimplementa-tionarevisibletoall.

Avendor’sdecisiontodisclosetechnologyisanirrevocablewaiverofitsabilitytoappropriatethere-turnsfromthattechnology.Theuseofhybridstrate-giessuggeststhatproprietaryvendorsareawareofthecompetitiverisksofsuchanappropriabilitywaiverandarethusexperimentingtoﬁndtherightcompromisebetweentotallyproprietaryplatforms(whichwouldberejectedbythemarket)andtotallyopenones(whichwouldeliminateallcompetitiveadvantage).Sofar,thetwohybridstrategieshavebeen(Fig.2):

•openingparts—wavingcontrolofcommoditylayer(s)oftheplatform,whileretainingfullcon-trolofotherlayersthatpresumablyprovidegreateropportunitiesfordifferentiation;

•partlyopen—disclosingtechnologyundersuchre-strictionsthatitprovidesvaluetocustomerswhilemakingitdifﬁcultforittobedirectlyemployedbycompetitors.

Theformerstrategyisimportantasanoffensivestrategytospeedadoptionofanewplatform-relatedstandardoraparticularimplementationofsucha

1280J.West/ResearchPolicy32(2003)1259–1285

Fig.2.Userrightsunderopenandquasi-opensourcelicenses.

standard.Waivingintellectualpropertyrightsmakesthestandard(orimplementation)moreattractivetocompetitorsandkeyusers,primingthepositive-feedbackbandwagoneffectsthatcanaccruetoearlymarketleaders.Italsoincreasesthenumberofprod-uctsthatareinteroperablewiththevendor’sproducts,particularlyimportantfornetworkingandothercom-municationsstandards.

Bothapproachesallowsophisticatedusers(suchaslargebusinessenterprises,universitiesorITindus-trysuppliers)tohelpimprovetheproductstheyuse,consistentwiththelong-identiﬁedroleoftechnicallyknowledgeableindustrialusers(vonHippel,1976).9.2.Futureplatformstrategies

Thestudyofstandardsandstandardsarchitecturecompetitionhasfocusedonthefollowingthreebasicapproaches:

•verticallyintegratedproprietarysystems,asrepre-sentedbytheIBM360(Chandler,1997;Moschella,1997);

•platformsassembledfromproprietarylayersthatarefreelylicensedtoall,suchas“Wintel”architecture(MorrisandFerguson,1993;Grove,1996);

•dejurestandardsnotsponsoredbyanysingleﬁrmbutsharedbyall,epitomizedby“opensystems”andEurope’sGSMdigitaltelephonestandard(Gabel,1987;FunkandMethé,2001).14Opensourcestandardsdifferfromotherunspon-soredopenstandardsmainlyindegree,totheextentthattheentryandimitationbarriersaredramati-callylower.Buttheideaofasharedstandard—withtheassociatedimplicationsforgovernanceanddifferentiation—isnotfundamentallydifferentbe-tweentheopensourceLinuxorFreeBSDanditsopensystems(Unixclone)ancestors.

Toalesserdegree,hybridplatformstrategieshaveexistedfordecades,drivenbytheever-increasingneedforsystemsinteroperabilitybetweenorwithinorganizations.Eventhemostproprietaryplatformin-corporatesopenindustrywidestandardssuchasISOcharactersets,ANSIC,EthernetorTCP/IP.Otherﬁrms(notablyIBMandMicrosoft)havetakenpor-tionsoftheirplatformsandgottenthemadoptedasindustrywidestandards.

Note,however,thatthenominallyopenGSMstandardizationeffortledbyNokiaandEricssonbuiltuponpatentportfoliosthatwereusedtoexcludeJapanesevendorsfromtheEuropeanmarket(Bekkersetal.,2002).

J.West/ResearchPolicy32(2003)1259–12851281

ThehybridstrategiesofSunandAppleblurthelinesbetweentheproprietaryandunsponsoredstandards.Byretaininganelementofcontrol,theyretainmanyofthecompetitivebeneﬁtsofsponsorship.However,byreducingduplicativeR&Dtheycancreatesharedcommunitiesthatinmanywaysareindistinguishableinpracticefromnominallyunsponsoredstandards—assumingthatthesponsorsmoveaggressivelyenoughtobuildsizablecommunitiesofadopter/collaborators.Theopensourcestrategiesstudiedalsocallatten-tiontotheuseofplatformextensionasastrategytodealwithcommoditizationoflower-levelplatformlayers.Forcomputerplatforms,suchextensionnor-mallyinvolvesdevelopingadditionalapplicationor“middleware”asthehighestlevelstandardslayersofanarchitecture.WhileApplehadfromthebeginningdifferentiateditsoperatingsystemthroughitseaseofuse,afterthatadvantagedisappearedin2001,ApplebeganbundlingconsumerapplicationstodifferentiateitsplatformamongPCbuyers.Microsoftbegansuchbundlingevenearlier,whenitincludeditsInternetExplorerapplicationwithitsWindows95operatingsystem.

Suchverticalintegrationintoapplicationssuggestsatleastapartialre-examinationoftheassumptionthatplatformssucceedthroughtheirabilitytoattractasupplyofthird-partyapplications.GallagherandPark(2002)haveshownthatin-houseapplicationsdevel-opmentwascrucialindecidingasuccessofplatformcontestsinthevideogameconsoleindustry.Ifthispatternismoregenerallyapplicable,thenitsuggestsﬁrmsneedtogarnertheﬁnancialresourcestosupplyacomplete(oratleastbasic)supplyofcomplemen-taryassetsexpectedbyadopters,ratherthanbuildingearlymarketshareperceptionstoattractthird-partysuppliersofsuchassets.

Inothercases,theattempttodifferentiatemaycon-tinuenottohigherarchitecturallayers,butwithsys-temintegrationordesign.InNovember2001,NokiaannouncedaplantolicenseitsPDA/mobilehandsetapplicationstorivals,eitherforusewiththeirownOS,ortorunonamulti-vendorOSdevelopedbytheSymbianjointventure(Nokia,2001).9.3.ImplicationsforopensourcedevelopmentHowopenisopenenough?Opensourceprovidesfewdirectbeneﬁtstothevastnumberofuserswho

lacktherequisitetechnicalskillstodotheirownde-velopment,butinsteadisbestsuitedfortechnicallyproﬁcientusers(suchasInternetserviceproviders)withstrongmotivations15forcustomization(WestandDedrick,2001).Thedegreetowhichopensourceaddsvaluebeyondthisnichedependsonhowmuchitenablesotherattributesmoredirectlyvaluedbyusers,suchasgreaterreliability,lowercostorexpandedva-rietyofcomplementaryassets.

Anexampleofsuchindirectbeneﬁtsistheprovi-sionofcomplementaryassets.Normally,theprovisionofapplicationsforanoperatingsystemiscontrolledbytheformal,publishedinterfaces(LangloisandRobertson,1995;WestandDedrick,2000).However,inanopensourcesystem,athird-partysoftwaresup-pliercanadditsowninterfacesasneededtoprovidefunctionalityunanticipatedbytheoriginalauthoroftheOS.ApplespeciﬁcallyopenedthelowestlayersofitsMacOSXarchitecturetoenableusersupportofunusualthird-partyhardwareconﬁgurations.

Anotherpostulatedindirectuserbeneﬁtofopensourcesystemsisincreasedreliabilitythroughthecon-currentdebuggingeffortsofawidelydistributedcom-munityofuser-programmers(Raymond,1999).Suchactivityappearedtobeanimportantgoalforallthreeﬁrmsstudied,astheproprietaryvendorssoughtvol-unteerlabortoﬁndandcorrectsomeofthegapsintheirsystems.KogutandMetiu(2001)reportthatmorethan70%oftheLinuxandApachecontributorshadmadeonlyonechangetothesourcecodetree.Thispatternisconsistentwithusersgettinginvolvedonlywhenthereisaproblemofgreatconcerntothem.Researchonlarge,successfulopensourceprojectssuchasLinuxandApacheassumesawayvarianceinwhatmaybeakeyindependentvariable:sizeoftheuser-programmercommunity.Anecdotalevidencesuggeststhatthemajorreasonopensourceprojectsfailisalackofuser-contributorstodothework.Inthiscase,sponsorsofopensourceprojects(whetherorganizationsorindividuals)vendorsfaceaparticu-larlyimportantadoptionchallenge:toattractenough

Directbeneﬁtfromaccesstoopensourcecodewouldappeartorequiresomecombinationofprogrammingskill,(personalorpro-fessional)motivationand(personalororganizational)slacktime.IntheirstudyofApacheusers,FrankeandvonHippel(2003)foundhighlyheterogeneousrequirementsamongusers(motivat-ingcustomization),andalsothatprofessionalskillpredictedcus-tomizationactivities.

1282J.West/ResearchPolicy32(2003)1259–1285

oftherightsortofusersearlyenoughtoimprovethequalityandfeaturesofthesoftware.16

Howapplicablearethesebeneﬁtstothehybridstrategies?Forthe“openingparts”strategy,musttheopenpartoftheplatformhavevalueonitsowntowinenoughuser-programmers?Orisitenoughthatitbepartofalargersystemofcrucialimportance,aswithApple’sreleaseoftheDarwinportionofitsmainstreamMacOSXoperatingsystem?

A“partly-open”strategy—suchasSun’sCommu-nitySourceLicenseorMicrosoft’sSharedSource—mustalsoattractattractinguser-programmers.Inthiscase,thequestioniswhetherthestrategyprovidesastableenoughallocationofthereturnsofinnovationbetweensoftwarevendorsandusers.Suchstrategiespresumethatthereisanintermediatelevelofdisclo-sureandgrantingofrights(betweenfullyopenandfullyproprietaryextremes)thatwillbevaluabletouserswithoutcompromisingtheIPRowners’competi-tiveconcerns.Onesuchintermediatepoint—theGNUPublicLicensewhichgrantsnearlyallrightsexcepttheuseincompetingproprietaryproducts—hashighvaluetousers17becauseofthecompleteandirrevocabledisclosure.Buttheotherformsusedthusfarseemtobediscountedbymostusers(particularlythosewithmorefullyopenoptions)andthusprovidelittleifanyadoptionincentive.9.4.Futureresearch

Thehybridopensourcestrategiesofproprietaryplatformvendorssuggestthatadditionalresearchisnecessaryonacrucialtrade-offintechnologicalinno-vation:resolvingtheconﬂictingimperativesofmak-16

IntheadoptiontypologyofRogers(1995),sponsorswouldwanttoattractinnovatorsastheuser-programmers,thenearlyadoptersasnon-programmingbetatesterswhowouldbetolerantofbugsandmissingfeatures,sothatacompletesolutioncouldbedevelopedsuitablefortheearlymajority.

17IrecognizethatadvocatesoftheGNU-stylelicensewouldarguethattheirlicenseismore“free”thantheBSDalternative,itthatguaranteesthatallderivativeswillalsobefree.However,suchperpetualfreedomisenforcedbyrestrictingtherightsgrantedwiththesoftware,whiletheBSD-stylelicensinghasfewifanyrestrictions.Aneconomicanalogisthechoicebetweengivingachildatrustfundorcash;theformerassuresthattheoriginalgoalwillalwaysbemet,whilethelattergivestherecipientfullrighttodoas(s)heseesﬁt.

inganinnovationsuccessfulandproﬁtingfromthatsuccess.

Inhisanalysisofthistrade-off,Teece(1986)fo-cusesonthecasewhereﬁrmsfaceweakappropriabil-ityregimesandthusmustsharethereturnsoftheirinnovation.Subsequentanalyseslargelyassumethatstrongappropriabilityisgoodandtobesought,whilelackofappropriabilityisalessdesirablecondition.Strategistsconsiderhowﬁrmsshouldsharetheeco-nomicsurplustheyhavegenerated,butnotthepossi-bilitythatﬁrmsshouldirrevocablyabandonsomeorallofthatsurplusbywaivingallfutureappropriabilityfromagiveninnovation.

However,proprietarysoftwareenjoysoneofthetightestappropriabilityregimesavailable,atleastincountriesthatenforceIPRlaws.Meanwhile,thethreedecadessincetheinventionofthemicroprocessorof-fernumerousexampleswhereinnovatorsenjoyedtightappropriability,butthelackofadoptionoftheirinno-vationmeantthattherewerenoreturnstoappropriate.TheconversionofproprietarysoftwaretoopensourcebyIBM,AppleandSun—forfeitingaportionoftheirIPR—suggeststhatthemoregeneralquestionofvoluntarilywaivingappropriabilitybearsfurtherstudy.Amongtheissuesthatmightbeconsidered:•IsforfeitingIPRmoreorlessdesirableincaseswherecompetitors(orsubstitutes)haveweakap-propriability?

•IssuchadifferentiationstrategymoreeffectivewhencompetingwithanorganizationwithtightIPRcontrol?Ordoesithavegeneralapplicabilityasastrategytopre-emptadoptionofcompetingtechnologies?

•Isittheeffectivenessdependentonthenatureofthetechnology,itsuse,ortheimportanceofcom-plementaryassets?Arethereanyotheradoptionprocesscharacteristicsrelevanttodecidingonsuchstrategies?

Otherpossibleareasforfuturestudyofopenstan-dardsinclude:

•Arethereanyconditionswhicharesufﬁcientforpredictingwhetheravendorshouldswitchfromproprietarytoopenstandards?Inparticular,willthepressureforopennessremainconﬁnedtolesssuc-cessfulvendors,orwilliteventuallyreachallﬁrms?•Aretherecategoriesofstandards(e.g.communica-tionsinfrastructure)thatwillalwaysbeopen?Are

J.West/ResearchPolicy32(2003)1259–12851283

thereany(e.g.documentinternalﬁleformats)thatwillalwaysbeclosed?

Researchonopensourcestrategiesbycommer-cialﬁrmsmightinclude:

•Underwhatconditionswillopenstandardsleadtothewidespreadadoptionofopensourceimplemen-tationsbycommercialvendors?

•

Howdurablearevariousapproaches(includingopeningparts,partlyopen)forcreatingbarrierstoimitationusingopensourcesoftware?Aresomelesslikelytocreatecompetitiveadvantagebecausetheydonotaddperceivedcustomervalue,orcon-verselyprovidetoomuchaidtocompetitors?

•

Isthedecisionbetweenopensourceandproprietarystrategiesamatterofmanagerialdiscretion?Oristhereanormative“bestpractice”forsimilarsituatedﬁrmsthatdictatestheappropriatechoiceforeachcontext?

•

Whenﬁrmsselecta“partly-open”strategy,whatistheappropriategovernanceforvendorsponsoredopensourcecommunities?Cantheconﬂictinggoalsofthetwopartiesbereconciled?

Acknowledgements

IwouldliketothankJasonDedrick,TinekeEgyedi,ScottEnsignandtwoanonymousreviewersforhelp-fulsuggestionsandcomments,aswellasseminarpar-ticipantsatCaseWestern’sWeatherheadSchoolofManagementandthe2002AcademyofManagementannualmeeting.Iamespeciallygratefultothespecialissueeditorsforadviceandencouragementthatwereinstrumentalindevelopingthepaper.MyappreciationgoestotherepresentativesofApple,IBMandSunwhogenerouslysharedtheirtimetobothexplainvar-iouscomplextechnologiesandalsotheinstitutionalrelationshipsbehindthosetechnologies.Theremain-ingerrorsareofcoursemyown.References

Anchordoguy,Marie,19.ComputersInc.HarvardUniversityPress,Cambridge,MA.

AppleComputer,1999.MacOSXServerEmbracesOpenSourceWithLaunchofDarwin.Pressrelease,16March:http://www.apple.com/pr/library/1999/mar/16opensource.html(accessed27December2001).

AppleComputer,2000.MacOSX:KernelEnvironment.AppleComputer,Cupertino,CA.

Arthur,W.B.,1996.Increasingreturnsandthenewworldofbusiness.HarvardBus.Rev.74(4),100–109.

Beggs,A.,Klemperer,P.,1992.Multi-periodcompetitionwithswitchingcosts.Econometrica60(3),651–666.

Bekkers,R.,Duysters,G.,Verspagen,B.,2002.Intellectualpropertyrights,strategictechnologyagreementsandmarketstructure.ThecaseofGSM.Res.Policy31(7),1141–1161.Besen,S.M.,Farrell,J.,1994.Choosinghowtocompete:strategiesandtacticsinstandardization.J.Econ.Perspect.8(2),117–131.Bresnahan,T.F.,Greenstein,S.,1999.Technologicalcompetitionandthestructureofthecomputerindustry.J.Ind.Econ.47(1),1–40.

Chandler,A.,1997.Thecomputerindustry:theﬁrsthalfcentury.In:Yofﬁe,D.B.(Ed.),CompetinginanAgeofDigitalConvergence.HarvardBusinessSchoolPress,Boston,pp.37–122.

Chposky,J.,Leonsis,T.,1988.BlueMagic:ThePeople,PowerandPoliticsBehindtheIBMPersonalComputer.FactsonFile,NewYork.

Cooper,C.,2001.IBM’sbigthinker,CNETNews.comhttp://news.cnet.com/news/0-1014-201-6947778-0.html,22August(accessed31December2001).

Dataquest,1996.WorldwidePCMarketGrew24%in1995.Pressrelease:http://www.gartnerweb.com/dq/static/about/press/pr-b9604.html,29January(accessed4January1997).

DiBona,C.,Ockman,S.,Stone,M.(Eds.),1999.OpenSources:VoicesfromtheOpenSourceRevolution.O’Reilly,Sebastopol,CA.

Eclipse.org,2002.EclipseProjectFAQ,19November:http://www.eclipse.org/eclipse/faq/eclipse-faq.html(accessed18December2002).

Egyedi,T.M.,2001.WhyJavaTMwasnotstandardizedtwice.Comput.Stand.Interf.23(4),253–265.

Eisenhardt,K.M.,19.Buildingtheoriesfromcasestudyresearch.Acad.Manage.Rev.14(4),532–550.

Ferguson,C.H.,Morris,C.R.,1993.ComputerWars:HowtheWestCanWininaPost-IBMWorld.TimesBooks,NewYork.Flamm,K.,1988.CreatingtheComputer:Government,Industry,andHighTechnology.BrookingsInstitution,Washington,DC.Franke,N.,vonHippel,E.,2003.SatisfyingHeterogeneousUserNeedsviaInnovationToolkits:TheCaseofApacheSecuritySoftware,Res.Pol.,thisissue.

Funk,J.L.,Methé,D.T.,2001.Market-andcommittee-basedmechanismsinthecreationanddiffusionofglobalindustrystandards:thecaseofmobilecommunication.Res.Policy30(4),5–610.

Gabel,H.L.,1987.Openstandardsincomputers:thecaseofX/OPEN.In:Gabel,H.L.(Ed.),ProductStandardizationandCompetitiveStrategy.North-Holland,Amsterdam.

Gabriel,R.P.,Joy,W.N.,1998.SunCommunitySourceLicensePrinciples:http://www.sun.com(accessed30December2001).Gallagher,S.,Park,S.H.,2002.Innovationandcompetitioninstandard-basedindustries:ahistoricalanalysisoftheUShomevideogamemarket.IEEETrans.Eng.Manage.49(1),67–82.Garr,D.,1999.IBMredux:LouGerstnerandtheBusinessTurnaroundoftheDecade.HarperBusiness,NewYork.

1284J.West/ResearchPolicy32(2003)1259–1285

Garud,R.,Kumaraswamy,A.,1993.Changingcompetitivedynamicsinnetworkindustries:anexplorationofSunMicrosystems’opensystemsstrategy.Strat.Manag.J.14(5),351–369.

Garud,R.,Jain,S.,Kumaraswamy,A.,2002.Institutionalentrepreneurshipinthesponsorshipofcommontechnologicalstandards:thecaseofSunMicrosystemsandJava.Acad.Manage.J.45(1),196–214.

Goff,Leslie,1999.SunandMicrosoftgopublic,Computerworld,20September1994.

Greenstein,S.M.,1997.Lock-inandthecostsofswitchingmainframecomputervendors:whatdobuyerssee?Ind.CorporateChange6(2),247–274.

Grove,A.S.,1996.OnlytheParanoidSurvive:HowtoExploittheCrisisPointsthatChallengeEveryCompanyandCareer.Doubleday,NewYork.

Hall,M.,2001.IBMroilsLinuxwaters,Computerworld,29October,pp.42–43.

Hess,R.,1995.AIX,MachTen,LinuxtoﬁllPPCUnixgap,MacWEEK,29May,p.16.

Katz,M.L.,Shapiro,C.,1985.Networkexternalities,competition,andcompatibility.Am.Econ.Rev.75(3),424–440.

Katz,M.L.,Shapiro,C.,1986.Technologyadoptioninthepresenceofnetworkexternalities.J.Pol.Econ.94(4),822–841.

Kim,E.E.,1999.TheJoyofUnix,LinuxMagazine,November:http://www.linux-mag.com/1999-11/toc.html(accessed30January2003).

Kogut,B.,Metiu,A.,2001.Open-sourcesoftwaredevelopmentanddistributedinnovation.OxfordRev.Econ.Policy17(2),248–2.

Kraemer,K.L.,Dedrick,J.,Yamashiro,S.,2000.Reﬁningandextendingthebusinessmodelwithinformationtechnology:DellComputerCorporation.Inform.Soc.16(1),5–21.

Langlois,R.,1992.Externaleconomiesandeconomicprogress:thecaseofthemicrocomputerindustry.Bus.HistoryRev.66(1),1–50.

Langlois,R.,Robertson,P.,1995.Firms,MarketsandEconomicChange,Routledge,London.

Leibovitch,E.,2001.Opensource’sblackhole:http://www.zdnet.com/enterprise/stories/main/0,10228,2714722,00.html,2May(accessed9October2001).

Liebowitz,S.J.,Margolis,S.E.,1999.Winners,LosersandMicrosoft:CompetitionandAntitrustinHighTechnology.IndependentInstitute,Oakland,CA.

Loukides,M.,1999.SomeThoughtsontheSunCommunitySourceLicense,http://java.oreilly.com/news/loukides0399.html(accessed30December2001).

MacCormack,A.,Herman,K.,1999.RedHatandtheLinuxRevolution.HarvardBusinessSchool,case#9-600-009,Boston,MA.

McKay,N.,1998.IBMbundlesApacheintoitsWebSphere,InfoWorld22June,p.3.

McKusick,K.,1999.TwentyyearsofBerkeleyUnix.In:DiBona,C.,Ockman,S.,Stone,M.(Eds.),OpenSources:VoicesfromtheOpenSourceRevolution.O’Reilly,Sebastopol,CA,pp.31–46.

Moltzen,E.F.,Burke,S.,1998.ApacheServerportedtoNT.ComputerResellerNews,16November,pp.5,10.

Moore,G.A,1991.CrossingtheChasm:MarketingandSellingTechnologyProductstoMainstreamCustomers.HarperBusiness,NewYork.

Moore,J.T.S.,2001.RevolutionOS,35mmﬁlm,WonderviewProductionsLLC.

Morris,C.R.,Ferguson,C.H.,1993.Howarchitecturewinstechnologywars.Harv.Bus.Rev.71(2),86–96.

Moschella,D.C.,1997.Wavesofpower:dynamicsofglobaltechnologyleadership,19–2010.AMACOM,NewYork.Mowery,D.C.,(Ed.),1996.TheInternationalComputerSoftwareIndustry:AComparativeStudyofIndustryEvolutionandStructure.OxfordUniversityPress,NewYork.

Nokia,2001.Nokiatolicenseamobileterminalsoftwareplatformandclientcomponentstohandsetvendors.Pressrelease,13November:press.nokia.com(accessed8April2002).

Raymond,E.S.,1999.TheCathedralandtheBazaar:MusingsonLinuxandOpenSourcebyanAccidentalRevolutionary.O’Reilly,Cambridge,MA.

Rogers,EverettM.,1995.DiffusionofInnovations,fourthed.,FreePress,NewYork.

Rosenberg,D.K.,2000.OpenSource:TheUnauthorizedWhitePapers.M&TBooks,FosterCity,CA.

Scannell,E.,Gardner,D.,1999.LinuxmeetsNetﬁnity.InfoWorld,2August,p.1.

Shankland,S.,1999.Applemodiﬁesopensourcelicense.CNETNews.comhttp://news.cnet.com/news/0-1003-200-341469.html,20April(accessed12June2001).

Shankland,S.,2002.MissionimpossibleatIBM?CNETNews.com,http://news.com.com/2008-1082-5391.html,30April(accessed18December2002).

Stallman,R.,1999.TheGNUOperatingSystemandtheFreeSoftwareMovement.In:DiBona,C.,Ockman,Stone,M.(Eds.),OpenSources:VoicesfromtheOpenSourceRevolution.O’Reilly,Sebastopol,CA,pp.53–70.

SunMicrosystems,2001.OpenOfﬁce.orgFAQ:http://www.sun.com/software/star/openofﬁce/faqs.html(accessed30December2001).

Taylor,M.,1996.ScottSays...‘KickButtandHaveFun’:ACandidInterviewwithScottMcNealy,Sun’sCEO:http://www.sun.com/960601/cover/(accessed30December2001).

Teece,D.,1986.Proﬁtingfromtechnologicalinnovation:implicationsforintegration,collaboration,licensingandpublicpolicy.Res.Policy15(6),285–305.

Torvalds,L.,Diamond,D.,2001.JustForFun:TheStoryofanAccidentalRevolutionary.HarperBusiness,NewYork.

vonBurg,U.,2001.TheTriumphofEthernet:TechnologicalCommunitiesandtheBattlefortheLANStandard.StanfordUniversityPress,Stanford,CA.

vonHippel,E.,1976.Thedominantroleofusersinthescientiﬁcinstrumentinnovationprocess.Res.Policy5(3),212–239.

Varhol,P.D.,1994.Trendsinoperatingsystemdesign:aninterviewwithLinusTorvalds,Dr.Dobb’sJournal,May,p.18.

Wayner,P.,2000.FreeForAll:HowLinuxandtheFreeSoftwareMovementUndercuttheHigh-techTitans.HarperBusiness,NewYork.

J.West/ResearchPolicy32(2003)1259–1285

West,J.,2000.AComparisonofPCStandardSwitchingDecisionsbyUSandJapaneseComputerUsers,Ph.D.dissertation.UniversityofCalifornia,Irvine,CA.

West,J.,2003.ThefallofaSiliconValleyicon:wasApplereallyBetamaxredux?In:Bettis,R.,(Ed.),StrategyinTransition.Wiley,NewYork,inpress.

West,J.,Dedrick,J.,2000.Innovationandcontrolinstandardsarchitectures:theriseandfallofJapan’sPC-98.Inform.Syst.Res.11(2),197–216.

1285

West,J.,Dedrick,J.,2001.Opensourcestandardization:theriseofLinuxinthenetworkera.Know.Technol.Policy14(2),88–112.

Wilcox,J.,2000.IBMtospend$1billiononLinuxin2001.CNETNews.com,http://news.cnet.com/news/0-1003-200-4111945.html,12December(accessed15December2000).

Zachary,G.P.,1991.FreeforAll:RichardStallmanisConsumedbytheFighttoEndCopyrightingofSoftware,WallStreetJ.,20May,R23.

因篇幅问题不能全部显示，请点此查看更多更全内容

查看全文