Do software firms collaborate or compete? A model of coopetition in community-initiated OSS projects

This is a sel -a chi ed e sion o an o iginal a icle. This e sion may di e om he o iginal in pagina ion and ypog aphic de ails. Au ho (s): Ti le: Yea : Ve sion: Copy igh : Righ s: Righ s u l: Please ci e he o iginal e sion: CC BY 4.0 h ps://c ea i ecommons.o g/licenses/by/4.0/ Do so wa e i ms collabo a e o compe e? A model o coope i ion in communi y- ini ia ed OSS p ojec s © he Au ho s, 2019. Published e sion Nguyen-Duc, Anh; C uzes, Daniela S.; Te je, Sna by; Ab ahamsson, Pekka Nguyen-Duc, A., C uzes, D. S., Te je, S., & Ab ahamsson, P. (2019). Do so wa e i ms collabo a e o compe e? A model o coope i ion in communi y-ini ia ed OSS p ojec s. e-In o ma ica, 13(1), 37-62. h ps://doi.o g/10.5277/e-In 190102 2019 e-In o ma ica So wa e Enginee ing Jou nal, Volume 13, Issue 1, 2019, pages: 37–62, DOI 10.5277/e-In 190102 Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s Anh Nguyen-Duc∗, Daniela S. C uzes∗∗ , Sna by Te je∗∗∗, Pekka Ab ahamsson∗∗∗∗ ∗Business school, Uni e si y o Sou h Eas e n No way ∗∗ Sin e Digi al ∗∗∗ Genus AS ∗∗∗∗ Uni e si y o Jy äskylä [email p o ec ed], [email p o ec ed], [email p o ec ed], [email p o ec ed] Abs ac Backg ound: An inc easing numbe o comme cial i ms a e pa icipa ing in Open Sou ce So wa e (OSS) p ojec s o educe hei de elopmen cos and inc ease echnical inno a i eness. When collabo a ing wi h o he i ms whose sough alues a e con lic s o in e es s, i ms may beha e uncoope a i ely leading o ha m ul impac s on he common goal. Aim: This s udy explo es how so wa e i ms bo h collabo a e and compe e in OSS p ojec s. Me hod: We adop ed a mixed esea ch me hod on h ee OSS p ojec s. Resul : We ound ha comme cial i ms pa icipa ing in communi y-ini ia ed OSS p ojec s collabo a e in a ious ways ac oss he o ganiza ional bounda ies. While mos o i ms con ibu e li le, a small numbe o i ms ha a e e y ac i e and accoun o la ge p opo ions o con ibu ions. We p oposed a concep ual model o explain o coope i ion among so wa e i ms in OSS p ojec s. The model shows wo aspec s o coope i ion can be managed a he same ime based on i m ga ekeepe s. Conclusion: Fi ms need o ope a ionalize hei coope i ion s a egies o maximize alue gained om pa icipa ing in OSS p ojec s. Keywo ds: COSS, coope i ion, collabo a ion, compe i ion, open sou ce so wa e, case s udy 1. In oduc ion Inc easingly, so wa e p oduc s a e no longe de eloped solely in-house, bu in a dis- ibu ed se ing, whe e de elope s collabo a e wi h “dis ibu ed collabo a o s” beyond hei i ms’ bounda y [1, 2]. This phenomenon in- cludes open sou ce so wa e (OSS) communi- ies, c owd-sou cing, and so wa e ecosys ems (SECO). This di e s om adi ional ou sou c- ing echniques in ha ini ia ing ac o s do no necessa ily own he so wa e de eloped by con- ibu ing ac o s and do no hi e he con ibu ing ac o s. Communi y-ini ia ed OSS p ojec s a e an example o he con ex in which ac o s coexis and coe ol e. F om i ms’ pe spec i e, i is bene icial o he de elopmen o so wa e p oduc s whose scopes exceeds hei own capabili ies by le e aging ex- e nal esou ces, explo ing oppo uni ies o en e new ma ke s [3], pe o ming an inside-ou p o- cess [4], and employing s a egic ec ui men s [5]. F om communi ies’ pe spec i e, he pa ic- ipa ion in such en i onmen p obably causes i ms o open up i s success ul p oduc s and p oduc lines o unc ional ex ensions by ex- e nal de elope s [1]. Ins ead o being exclusi e and localizing p oduc de elopmen , i ms a e Submi ed: 1 May 2018; Re ised: 24 June 2018; Accep ed: 31 July 2018; A ailable online: 1 Oc obe 2018 38 Anh Nguyen-Duc e al. explo ing di e en ways o in i e con ibu ions om ex e nal ac o s wi hou e ealing co e ech- nology, business alue and cus ome ela ion- ships [6]. Be o e he ull po en ial ad an ages o open sou cing a e le e aged, comme cial i ms need o conside se e al conce ns. A he o ganiza- ional le el, he i m’s bene i and he commu- ni y goals a e no always he same [7]. Pa ici- pa ion o comme cial i ms in OSS p ojec s wi h hei di e se mo i a ions and business s a e- gies migh in oduce a iance, and some imes con lic s in p ojec e olu ion [3]. Exis ing e- sea ch on OSS highligh s he ole o collabo- a ion wi h ex ensi e esea ch on communica- ion and coo dina ion p ac ices, pa e ns and lessons lea n om OSS communi ies [8 – 11]. How- e e , he e seems o be a less esea ch conce ns abou he con lic s among i ms ega ding o hei s a egic de elopmen . Fi ms a emp o gain compe i i e ad an ages om hei pa ic- ipa ion in OSS p ojec s [12]. When he e oc- cu misma ches in e m o in e es s and objec- i es, i ms may beha e uncoope a i ely in o - de o p e en o he s om achie ing hei goals [13]. The con lic occu s no only a he man- age ial le el, such as p ojec go e nance [14], bu also a he ope a ional le el, such as code con ibu ion, bug ixes, and equi emen elici a- ion [3,15–17]. Coope i ion, as a business phenomenon, is abou collabo a ing and handling a i m’s com- pe i i e ad an ages when pa icipa ing in OSS p ojec s [18,19]. In a coope i i e en i onmen , i ms coope a e wi h each o he o each a highe alue c ea ion compa ed o he alue c ea ed wi hou he in e ac ion. The basic assump ion o coope i i e ela ionships is ha all ac i i ies should aim a he es ablishmen o a bene icial pa ne ship wi h o he i ms, including pa ne s who may be conside ed as a kind o compe i- o [20]. Since coope i ion applies o in e - i m ela ionships, OSS p ojec o e s an ideal con- ex o unde s anding he phenomenon among i ms ha de elop and u ilize a common so wa e codebase [16]. Empi ical esea ch on coope i ion is sca ce, especially s udies in So wa e Enginee ing (SE) and a he o ganiza ional le el [13]. Resea ch in his a ea is p obably hidden by he inconsis en ea men o he c oss-disciplina y na u es o coope a ion and compe i ion, and hei ela ed cons uc s. Ou esea ch objec i e is o explo e how i ms in e ac and manage he phenomenon o coope i ion in OSS p ojec s. To bes o ou knowledge, he e exis s only a ew s udies ha examine he phenomenon o coope i ion among comme cial i ms in OSS p ojec s [3,13,15,17]. Resea ch ques ions (RQs) we e de i ed om his esea ch objec i e. Fi s ly, we aimed a unde - s anding he basic ounda ion on i m pa icipa- ion in OSS p ojec s. Based on his knowledge, we explo ed u he heo e ical elemen s o coope- i ion. We use he e he wo d “coope i i ely” as an ad e b o coope i ion: – RQ1: How do comme cial i ms pa icipa e in communi y-ini ia ed OSS p ojec s? – RQ2: How do comme cial i ms manage coope i ion wi h o he i ms in such con ex ? Ou con ibu ions a e wo olds, i s ly we po - ayed he si ua ions whe e bo h compe i ion and collabo a ion occu s in OSS p ojec s. Con- side ing he body o knowledge abou i m pa ic- ipa ion in OSS p ojec s, ou wo k con i ms some pa e ns and also ex ends hem by explo ing he i m awa eness, coope i ion and hei an eceden ac o s. Adop ing a mixed-me hod esea ch, we quan i a i ely examine o ganiza ional in e ac- ion pa e ns and quali a i ely explo e how i ms pe cei e and employ coope i ion s a egies. Sec- ondly, we heo ize cons uc s o coope i ion by p oposing a Coope i ion in Open Sou ce So wa e (COSS) model. P e ious s udies ha men ion he e m “coope i ion” [3,15], do no in es iga e he cons uc s unde his phenomenon. Hence, o ou bes knowledge, his is among he i s s udies in SE in es iga ing his concep . The p oposed model e eals building blocks o coope i ion in OSS i ms ne wo k and i s ela ionship o conse- quen ac o s. The s udy is o ganized as ollows: Sec ion 2 p esen s a backg ound abou coope i ion and i m pa icipa ion in OSS p ojec s. Sec ion 3 desc ibes ou esea ch me hodology, Sec ion 4 p esen s ou indings, and Sec ion 5 discusses he indings. Finally, Sec ion 6 concludes he pape . Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 39 2. Backg ound and ela ed wo k 2.1. The phenomenon o coope i ion The o igin o coope i ion is om business e- sea ch when in es iga ing buye and selle ela- ionships wi hin a business ne wo k [18,19]. The ade-o be ween coope a ion and compe i ion is emphasized as a mean o c ea ing a p og ess among ac o s in ol ed in long- e m ela ion- ships. Coope i ion concep ualizes he in e ac ion among i ms in ela ion o hei s a egic de el- opmen [18, 19]. Dagnino e al. de ined coope- i ion as “a kind o in e - i m s a egy which consen s he compe ing i ms in ol ed o manage a pa ially con e gen in e es and goal s uc u e and o c ea e alue by means o coope i i e ad- an age” [21]. The au ho s p oposed wo o ms o coope i ion, a dyadic coope i ion (conce ns among wo- i m ela ionships) and a ne wo k coope i ion (in ol ing mo e han wo i ms, i.e. alue chain) [21]. Beng sson a gued ha a dyadic ela ionship is a pa adox ha eme ges when wo i ms coope a e in some ac i i ies, such as in a s a egic alliance, and a he same ime com- pe e wi h each o he in o he ac i i ies [19]. I means ha ac o s wi hin a i m need o be di- ided o ake cha ge o ei he collabo a ion o compe i ion. Coope i ion can occu in a mo e complex o m, wi h a ne wo k o i ms. The coope i- ion s a egy can be applied a a mic o le el (among unc ional and di isional depa men s in a i m), a meso le el (among i ms in he same indus y, be ween endo and supplie ) and a mac o le el (among clus e o i ms o i ms ac oss indus ies) [21]. Li e a u e also dis- cusses some an eceden ac o s ela ing o coope- i ion a he mic o le el, such as sha ed i- sion, pe cei ed us and pe cei ed bene i s [22]. A s udy poin s ou some possible impac s o coope i ion on knowledge sha ing and job/ ask e ec i eness [22]. By selec ing a highly inno- a i e OSS p ojec ha con ibu es o i ms’ s a egic alues, we illus a e dependencies be- ween compe i o s due o s uc u al condi ions, why and how compe i o s coope a e. 2.2. Collabo a ion in OSS p ojec s Collabo a ion is an aspec o coope i ion ha is much explo ed in OSS p ojec s. I is com- mon o look a OSS p ojec s’ a chi es o e eal communica ion, collabo a ion and coo dina ion app oaches, equency, pa e ns and bes p ac- ices a di e en le el o analysis [2,23 – 31]. Ea ly esea ch has obse ed an onion-like s uc u e o con ibu ion in OSS p ojec s [24 – 27]. A he cen e o he onion a e he co e de elope s, who con ibu e mos o he code and ake ca e o he design and e olu ion o he p ojec . In he nex ing ou a e he co-de elope s who submi pa ches (e.g. bug ixes), which a e e iewed and checked in by co e de elope s [28]. Fu he ou a e he ac i e use s who do no con ibu e code bu p o ide use-cases and bug- epo s as well as es ing new eleases. The awa eness o peo- ple and ac i i ies h ough OSS social s uc u es enhances collabo a ion e ec i eness and ensu es ha li le e o is was ed in duplica e wo k [30]. A la ge amoun o s udies in es iga es he com- bina ion o social and echnical aspec s o OSS p ojec s, by analyzing a social ne wo k c ea ed by con ibu o s who wo k and communica e in he same se o iles [32 – 35]. Bi d e al. [34] showed ha a socio- echnical ne wo k o so - wa e modules and de elope s is able o p edic so wa e ailu e p oneness wi h g ea e accu acy han o he p edic ion me hods. Wol e al. [35] o med a de elope - ask ne wo k o explo e he impac o de elope communica ion on so wa e build in eg a ion ail. A common assump ion o hese s udies is ha de elope s beha e ega dless o hei comme cial a ilia ions in OSS p ojec s, indica ing by unweigh ed analysis app oaches when o mula ing he social ne wo ks. In case a signi ican numbe o de elope s om i ms con ibu es o he p ojec , o ganiza ional ea- u es, such as i ms’ s a egies and go e nance mechanism migh in luence he communica ion s uc u es o he OSS p ojec s. In his wo k, we will use he social ne wo k analysis (SNA) o in- es iga e in e ac ion pa e ns, i.e. collabo a ion and compe i ion in OSS p ojec s. While we also o m he de elope - ask-de elope ne wo k, he 40 Anh Nguyen-Duc e al. di e ence is ha he ela ionship is analyzed a i m le el. 2.3. Collabo a ion in OSS p ojec s A heo e ical model links heo e ical elemen s in a ce ain seman ic manne , i.e. a causal ela- ionship, helping o design da a collec ion and analysis. Li e a u e e eals ac o s ha lead o he occu ence o collabo a ion and compe i ion (an eceden ac o s), and hei impac on i ms’ ou comes (consequen ac o s). I is no ed ha we do no aim o model comple eness, bu o a ounda ion o u he in es iga ion. The u he in es iga ion would disco e which ac o s alid in he con ex o so wa e indus y, pa icula ly OSS p ojec s. As seen in Figu e 1, coope i ion is he s ud- ied cons uc , and i is linked o i s an eceden ac o s, i.e. s uc u al condi ion, s a egic ision, us and pe cei ed bene i s [22,36–41]. S a egic ision : sha ing s a egic ision is essen ial o coope a ion a eam le el [22] [35], as he ision e lec s impo an ag eemen s o belie s and assump ions ha consequen ly b ing in e nal s abili y o he coope a i e a i ude [36]. A he s a egic le el, ision ypically is abou he i m’s alue and business de elopmen . Sha ed ision d aws a oadmap o he o ganiza ion o i m, se ing he p io i ies o hei eam plan- ning and implying i s c i ical de e minan ole in lessening malign compe i ion [22]. The ision can be sha ed ia mee ings o wo kshop wi h high-le el manage s. T us : is conside ed as a ela ionship o e- liance among membe s o a eam o an o gani- za ion. T us is de ined as “ he willingness o a pa y o be ulne able o he ac ions o an- o he pa y based on he expec a ion ha he o he will pe o m a pa icula ac ion impo - an o he us o , i espec i e o he abili y o moni o o con ol ha o he pa y?” [42]. The impo ance o us in he success o in e pe - sonal ela ionships is epo ed p e iously in OSS p ojec s [37, 38]. Mo eo e , us is he key o ans o ming OSS as a communi y o indi idual de elope s, o OSS as a communi y o i ms [39]. The coope a ion ha cap u es he le el o coo - dina ed ac ions be ween eam membe s in hei e o s o achie e mu ual goals canno be ealized wi hou us among he membe s. Pe cei ed bene i : on one hand, pe cei ed bene i s a e associa ed wi h a coope a i e a - i ude, in ol ing compa ible in e es s as com- mon bene i s can mo i a e collabo a ion, le e age eam o pe son’s capabili ies o ob aining such bene i [40]. In OSS p ojec s, pe cei ed bene i s o pa icipa ing in he communi ies a e educed de elopmen cos , communi y knowledge, and educed main enance cos . On he o he hand, pe cei ed bene i is also associa ed wi h a com- pe i i e a i ude. Indi iduals a e likely o pu - sue hei own objec i e a he expense o e all eam’s goal [41]. This could be applicable o o ganiza ion in an ecosys ems o supply chains. The mo e bene i a i m pe cei e o ob aining a con lic ing a i ac o esou ce, he mo e hey likely o compe e o e he esou ce [22]. In ou heo e ical amewo k, coope i ion is also associa ed o i s consequen ac o s, i.e. knowledge sha ing and ask e ec i eness [22,43]. Knowledge sha ing a o ganiza ional le - els is seen as sha ing o o ganiza ional expe ience and knowledge, i.e. echnical know-how, domain expe ise, wo k p ac ice, e c wi h o he collabo a- o s, and hence inc easing he o e all knowledge in he join p ojec [22]. As knowledge is a c i ical sou ce o compe i i eness, managing knowledge sha ing among membe s o an o ganiza ion plays a p ominen ole in sus ainable compe i i e ad- an age [43]. Task e ec i eness in eam collabo a ion ep esen s indi iduals’ pe cei ed capaci y o con- duc ing collabo a i e asks, whe eas knowledge sha ing enhances he abili y o collabo a o ’s knowledge exchange. 3. Resea ch app oach 3.1. S udy design We conduc ed his wo k by using a wo-phase mul iple-case s udy design [44]. The phases in he esea ch occu due o he disc e e con inua- ion o ou in e nal esea ch p ojec . Compa ed Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 41 Figu e 1. A heo e ical amewo k o coope i ion (adap ed om [22]) o desc ip i e and con i ma i e case s udies, ex- plo a o y case s udies a e sui able o he i s phase esea ch as we would like o disco e he phenomenon o coope i ion, whe he i exis s, in which o m and i s ela ionship o i s con- ex se ing. This phase was done as a pa o a mas e hesis. In he second phase, we con- duc ed a desc ip i e s udy on desc ibing collab- o a ion, compe i ion in he selec ed cases. In he hi d phase, we ound ano he case s udy o con i m he quali a i e indings. This s ep was conduc ed o alida e wha we obse ed in he i s wo cases. We ollowed he guide- line by Runeson and Hös [45] o execu e case s udy, including case selec ion, da a collec ion and analysis. Case selec ion is no s aigh o wa d. The e a e abundan OSS p ojec s a ailable; many o hem a e abandoned o indi idual e o s. A b ains o m session was conduc ed among he pape ’s au ho s o decide case selec ion c i e ia as below: – Comme cial pa icipa ion: he OSS p ojec should ha e mul iple comme cial i ms pa - icipa ing in he de elopmen . In addi ion, he e mus be an adequa e way o iden i y hem. – Success ul and on-going: he OSS p ojec mus be success ul and on-going. This implies ha he p ojec a ac s de elope s and he de elopmen o he so wa e is p og essing. – Ac i e p ojec s wi h many ac i i ies: he OSS p ojec mus ha e a high le el o communica- ion and code commi s in he p ojec , showing by ich da a a chi e. By e iewing li e a u e on OSS p ojec s in SE, we lea n se e al OSS p ojec s ha we e com- monly in es iga ed in SE esea ch, such as Apache, Mozilla, Eclipse and Linux [46]. The selec ed cases should no only sa is y he selec ion c i e ia, bu also no el in SE esea ch. We we e sugges ed o Wi esha k by a colleague who pa icipa ed in he p ojec . Many easons con ibu ed o his choice. Fi s ly, he con ibu o lis and communi y ac i - i y e ealed high pa icipa ion and in ol emen o comme cial companies. Wi esha k is a ypical ins ance o a OSS p ojec . The p ojec uses so - wa e in o malisms o de elopmen collabo a ion, he de elope s a e a mix o i m-paid de elope s and olun ee s, and he so wa e is licensed unde he GNU Gene al Public License (GNU GPL). Wi esha k is also a e y success ul on-going OSS p ojec , wi h a high numbe o con ibu o s and ac i e use s, consis en ly pushing de elopmen o wa d. Ha ing selec ed Wi esha k as he i s case, we p oceeded o ind and selec he second case o ou s udy. To be able o do a li e al eplica ion, he second case should ha e simila p ope ies as he i s case. A e a long pe iod o sea ching, we ended up wi h h ee p omising cases ha ma ched he speci ica ions: Ho de, Samba and Wine. F om he compa ison i was e iden ha Samba was e y simila o Wi esha k, i.e. bo h p ojec swe e licensedunde GNUGPL,bo h p ojec s had many i ms pa icipa ing, and hey bo h had a yea ly con e ence whe e de elope s cane oge he o discuss u he de elopmen and socialize. We planned o ha e he hi d case o alida e he quali a i e indings om Wi esha k and Samba. Among se e al OSS p ojec s we 42 Anh Nguyen-Duc e al. Figu e 2. O e iew o he esea ch p ocess a emp ed o con ac , Boo s ap de elope s we e he one ag eed o pa icipa e in he s udy. The esea ch p ocess is desc ibed in Figu e 2. A he p e-s udy phase, li e a u e e iew and b ains o ming wi h expe s we e done o come up wi h esea ch objec i e and s udy design. A he explo a o y and desc ip i e phase, he i s wo cases we e in es iga ed o unde s anding how comme cial i ms pa icipa ed in OSS p ojec s, i he phenomenon coope i ion exis s and in which o m. As he explo a i e na u e o his phase, a wide ange o opics was disco e ed, such as collabo a ion pa e ns, i m awa eness, compe i- ion, code p ac ices, e c. The da a we e ex ac ed om p ojec a chi e, i.e mailing lis s, bug ack- ing sys em and code eposi o y. In his phase, we also collec ed quali a i e da a, i.e. in e iewing ele an s akeholde s o explo e in-dep h phe- nomenon obse ed om he quan i a i e da a. A he con i ma i e phase, we conduc ed some in e iews o con i m and o alida e he obse - a ion om he i s wo cases. 3.2. Case desc ip ion Wi esha k 1 is an OSS oolki de eloped by a com- muni y o ne wo king expe s a ound he wo ld unde he GNU Gene al Public License. The p ojec is o icially ope a ed unde he Wi esha k name since May 2006. Ou o he 802 de elope s lis ed in Wi esha k con ibu o lis , 342 we e classi ied as i m-paid de elope s (43%). The e- maining 460 de elope s (57%) we e classi ied as olun ee ing de elope s. The i m-paid con ibu- ions come om 228 i ms. Samba 2 is an OSS sui e ha p o ides ile, p in and au hen ica ion se ices o all clien s using he SMB/CIFS p o ocol. Samba is licensed unde he GNU Gene al Public License, and he Samba p ojec is a membe o he So wa e F eedom Conse ancy. In Samba, 316 de elope s we e e alua ed, whe e 182 (57%) o hem we e classi ied as i m-paid de elope s. The con ibu- ions come om 45 i ms. Communica ion and collabo a ion be ween de elope s in he Wi e- sha k and Samba communi y mainly occu in wo places; he de elope mailing lis and he bug acking sys em. La e , a hi d OSS p ojec was selec ed as a mo e ecen p ojec o p o ide complemen- a y quali a i e da a. Boo s ap 3 is a on end Ja asc ip -based amewo k o de eloping e- sponsi e, mobile i s p ojec s on he web. The p ojec was eleased as an OSS p ojec since 2011 unde MIT license. Boo s ap we e con ibu ed by la ge i ms, such as Twi e and Gi Hub. A he ime he esea ch was conduc ed, Boo s ap has been he mos -s a ed p ojec on Gi Hub, wi h o e 90.000 s a s and mo e han 38.000 o ks. The communica ion in Boo s ap was done ia 1h ps://www.wi esha k.o g 2h ps://www.samba.o g 3h p://ge boo s ap.com Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 43 many channels, i.e. S ackO e low, Slack, and Gi Hub acke . Sou ce code and issue manage- men was done ia Gi Hub. 3.3. Da a collec ion 3.3.1. Quan i a i e da a The main sou ce o quan i a i e da a is om mailing lis s, code and issue eposi o ies, as hey a e common da a sou ces when s udying OSS [8,15,23,47]. We collec ed h ee ypes o da a, namely de elope p o ile, i m p o ile and commu- nica ion da a. The de elope p o ile was ound om p ojec public pages, such as p ojec wiki and con luence page. Basic in o ma ion, like de- elope s’ email add esses and imes amp o ile commi s we e ex ac ed om JIRA and GIT. F om de elope s’ p o iles, we we e also able o iden i y he lis o i ms in a OSS p ojec . An in i a ion o in e iew was sen in a snowballing manne . A e i m-paid de elope s accep ed ou in i a ion o in e iew, basic in o ma ion abou he i m was equi ed by us. Besides, i m in o - ma ion was also collec ed om online sou ces, such as company websi e, and published ma e i- als. The communica ion da a was collec ed om wo main sou ces, namely issue acking sys em and mailing lis . These sou ces con ained de ailed in o ma ion abou e en s and ac i i ies ha had occu ed in he communi ies se e al yea s back in ime. Table 1 gi es an o e iew o when he sou ces we e i s used and how many en ies hey ha e oday in Wi esha k and Samba. 3.3.2. Iden i ica ion o i m pa icipa ion In o ma ion whe he a pa icipan is a i m-paid o olun ee de elope , is no gene ally a ail- able in OSS p ojec s. Consequen ly, we needed o come up wi h a classi ica ion echnique o iden i y i ms’ pa icipa ion. The app oach has been success ully used in a p e ious s udy [48]. The ollowing in o ma ion was e alua ed in he p ocess o classi ying he de elope s: – Cu en s a us in he communi y: ac i e o no any mo e. – Email domain: The email domain used by a de elope can e eal i m associa ion. We ega d i as unlikely ha a de elope use a job email o pa icipa e in an OSS p ojec i i is no ela ed o he job as a paid de- elope . This measu e is he mos dis inc i e classi ica ion en i y. – Email signa u e: Some de elope s ha e hei employmen i m name as pa o hei email signa u e, which hey use when pos ing o he mailing lis o bug acke . – Pe sonal homepage: Sea ching o a de el- ope ’s name on he web can gi e di ec ions o a pe sonal homepage o blog ha migh e eal company associa ion. – Social ne wo ks: Sea ching o a de elope ’s name on social ne wo ks like LinkedIn and o he p o essional pages migh e eal i m a ilia ion. – P esen a ions and con e ences: De elope s ha gi e p esen a ions commonly include name and i m in he p esen a ion slides, which a e easy o ind by a web sea ch. Some issues we e aced when iden i ying con- ibu o s’ a ilia ions. Fi s ly, he e is a di e en le el o con ibu ions in OSS p ojec s. The e is o en a lack o in o ma ion abou wha is equi ed o become a con ibu o . Mo eo e , majo i y o he pa icipan s in he mailing lis only pos ed one mail, which makes i a was e o ime and e o o iden i y hese pa icipan s as he con ibu ion owa ds he i m’s in e ac ion and so wa e de el- opmen is minuscule. We decided o exclude de el- ope s wi h less han en en ies in he mailing lis o bug acking sys em. Secondly, ma ching name, alias and email add ess is no always s aigh o - wa d. In Wi esha k, he spam p o ec ion policy hides he ull email add ess, o ins ance: “F om: [de elope name] <name@xxxxxxxxxxxxx>”. Mo eo e , en ies in he bug acking sys em ha e email lis ed, bu no name. The code eposi o y en ies in Wi esha k does no con ain name o mail o he de elope , ins ead a use name o a nick- name is used. We had o use p ojec wiki pages and pe sonal con ac s wi h some co e de elope s o he p ojec o p o ide mapping o mos o he use names o he ac ual de elope s. 44 Anh Nguyen-Duc e al. Table 1. Summa y o quan i a i e da a om Wi esha k and Samba P ojec Da a sou ce Da e o i s en y # o en ies Wi esha k Mailing lis 31.05.2006 27230 Bug acking sys em 08.04.2005 7862 Code eposi o y 16.09.1998 42794 Samba Mailing lis 03.01.1997 90588 Bug acking sys em 24.04.2003 9659 Code eposi o y 04.05.1996 84699 3.3.3. Quali a i e da a Rega ding o quali a i e da a, in e iews we e selec ed om a con enien sample consis ing o he i m-paid de elope s om Wi esha k, Samba and Boo s ap. Ten in e iews we e conduc ed as seen om Table 2. In Wi esha k and Samba, we managed o ha e in e iews om i ms in a co e laye and a pe iphe al laye (de ail as shonw in Figu e 6). Due o non-disclosu e ag eemen s, we did no e eal he ac ual iden i y o companies (quan i a i e da a was publicly a ailable, hence did no ha e his cons ain ). We used alias D1 o D10 o ep esen o such i ms. As we did no know much abou he popula- ion, we aimed o a non-p obabilis ic sampling echnique using a conjunc ion o pu posi e and snowball sampling. In Wi esha k, we used an ex- is ing connec ion o one o he co e con ibu o s as a s a ing poin , and asked o sugges ion o de elope s ha could be in e es ing o in e iew nex . The co e con ibu o poin ed ou ele an de elope s o he esea ch opic, and assis ed in con ac ing hem by pos ing ou in e iew in- i a ion on he co e con ibu o mailing lis . In Samba, we selec ed ele an de elope s in he OSS p ojec based on he quan i a i e da a and sen in e iew in i a ions o hese by email. In Boo s ap, we had a de elope ac i ely con ibu - ing o he p ojec in ou pe sonal ne wo k. F om him, we go wo mo e in e iews wi h i m-paid pa icipan s in Boo s ap. The in e iew guide consis ed o bo h closed and open ques ions. The closed ques ions we e mainly used in he in oduc ion phase o he in- e iew o solici backg ound in o ma ion abou he esponden , i m and OSS p ojec con ex . In addi ion, he closed ques ions we e used o con i m o a ibu e s a emen s gi en by o he de elope s. The open ques ions we e used o collec in o ma- ion abou : (1) wo k p ocess/b idge enginee ole, (2) i m awa eness/o ganiza ional bounda y and (3) posi ion in he communi y/con ibu ions. The in e iew guide and in e iew ques ions is publicly a ailable. The in e iews we e conduc ed in En- glish, excep o one in No wegian. The du a ion o he in e iews anged om 45 minu es o 72 min- u es. All he in e iews we e eco ded o acili a e subsequen analysis and minimize po en ial da a loss due o no e- aking. These eco dings we e he ea e ansc ibed e ba im. T ansc ibing audio eco ds esul ed in 55 pages o ich ex . 3.4. Da a analysis 3.4.1. Social ne wo k analysis (SNA) SNA is a common app oach o in es iga e com- munica ion and collabo a ion pa e ns based on da a om mailing lis s o issue acking sys ems [32 – 35,49]. This has been ex ensi ely used o cons uc ing a de elope - ask ne wo k and mea- su ing di e en ne wo k ea u es [32 – 35]. We adop ed his app oach in i m le el, o unde - s and he collabo a ion pa e n among i ms ia communica ion ne wo ks. Consequen ly, we used he i ms as nodes and he in e ac ion be ween i ms as edges. In e ac ion among i ms is ep e- sen ed by communica ion ia ei he a mailing lis o commen s on an issue acking sys em. The SNA was done in ou s eps: – S ep 1: Cons uc discussion ees om a mail- ing lis and an issue acking sys em. A dis- cussion ee consis s o an iden i ie node, a sou ce node and a se o esponde nodes (which can ange om none o many). The de elope ha ini ia es a discussion is e- ga ded as he sou ce, and he de elope s ha Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 51 had used di ec and/o p i a e communica ion channels o asking o help om he domain expe s in he p ojec . Communica ion channels used a e e-mail and ins an messaging, Skype and elephone. D3 said: “I ha e done i [con ac ed de elope s di ec ly] some imes in he pas . No jus as a gene al I am s uck, can you help, bu because i would be an a ea I knew he o he guy was wo king on.” The p i a e communica ion is usually he esul om a g adual es ablishmen ia public communica ion, as men ioned by D6: “Usually I end o do R&D asks mysel . I o en seek o e iews o my wo k. When I need he assis ance, I will go di ec ly o a de elope in he communi y.” Compa ing o public channels, D8 consid- e ed p i a e communica ion as a way o es ablish high-quali y con ac poin s and po en ial collab- o a ion o u he p ojec s. He men ioned ha a o k om p ojec mains eam should p oba- bly include bes de elope s in he communi y who a e no necessa ily he guy in he “onion co e”. I is also s a ed ha a p i a e channel is a quick and e icien communica ion medium. D9 explained ha he used ins an messaging o con ac ing de elope s in he communi y when he wan ed a quick eedback. P i a e communi- ca ion seems o be in a o compa ing o public communica ion. D9 men ioned: “We y o ad- d ess as much as we can he issues ha come o us .. . No mally i we ge a p i a e message abou an issue, we will ake i wi h highe p io - i y” D5 men ioned ha when discussing legal o secu i y sensi i e issues, he used a p i a e com- munica ion channel. The na u e o such issues in okes he use o p i a e channels as pos ing i in he public channels may esul in secu i y b eaches o simila ly bad si ua ions. Al hough none o he o he de elope s said any hing abou he use o di ec channels o such issues, we belie e ha i is a common p ocedu e in mos OSS p ojec s. 5.3. T us T us is one o he undamen al ai s o a success- ul collabo a i e en i onmen [29,51 – 53]. Ray- mond s a ed ha “open-sou ce cul u e has an elabo a e se o cus oms?[which] egula e who can modi y so wa e, he ci cums ances unde which i can be modi ied, and (especially) who has he igh o edis ibu e modi ied e sions back o he communi y” [54]. In ou cases, in e - iewee s a ed ha he success o OSS p ojec s is meaning ul o hem. Fo ins ance, wi h he ad ance o he Wi esha k ool, D4 can use i o se e o his daily wo k. Based on us , de- elope s can collabo a e o he sake o hei OSS p ojec . D3 said ha hey ha e con ac ed us able de elope s di ec ly o a oid asking silly o dumb ques ions in public: “I go ela ionships wi h o he de elope s and some imes we don’ wan o ask in mailing lis causes i is a eally s upid ques ion and you do no wan o ask he whole mailing lis , so you jus ask he guy you us ”. When a de elope needs help o design a code o ix a bug, o he de elope s would be willing o assis . By helping one ano he , de el- ope s demons a e hei skills and knowledge, which de elops a posi i e expec a ion o com- pe ence and eliabili y. Le el o us is ela ed o he s a us o he de elope s in OSS p ojec s, which is e iden in he ollowing sec ion. The obse a ion is aligned wi h p e ious esea ch on he ole o us in success ul in e pe sonal ela- ionships [37,38]. 5.4. Pe cei ed bene i s Despi e he isks associa ed wi h compe i o s, many i ms decided o be open in sha ing and synch onizing hei sou ce code wi h OSS com- muni ies. Sou ce code can be synch onized wi h ups eam de elopmen in OSS p ojec s, o in- s ance, desc ibed by D5: “In gene al, ou phi- losophy is o de elop ups eam i s and hen back-po changes ha ha e been app o ed by he ups eam communi y in o ou p oduc s. We s ay e y in ol ed in he communi ies and y o keep he di e ences be ween ou packaged so wa e and ups eam so wa e o he mini- mum necessa y.” Fi ms pe cei e bene i s wi h such in ol emen as a oiding main enance and me ging issues when combining public pa s o p i a e pa s o sou ce codes. D10 illus a ed o his idea: “.. . i you a e o make a change 52 Anh Nguyen-Duc e al. Figu e 9. The ole o ga ekeepe in a comme cial i m in he co e, and you wan o keep i p i a e, you will ha e o o k he p ojec and main ain i you sel . (...) I belie e, in he gene al case, ha you gain mo e om con ibu ing o he de- elopmen , ha e aining you code om he communi y”. D1 men ioned ha “We do no ha e o main ain ou own code base and syn- ch onize i . We jus commi code o he sou ce and ha e i he e. I we had no had he commi access as easy as I do, we could ha e had ou own e sion o Wi esha k and he sou ces, bu hen we would ha e o do mo e wo k in me g- ing ou e sion wi h he new eleases o Wi e- sha k.” Fi ms also conce n abou hei social posi- ions in he p ojec s. I is appa en ha a cen al posi ion in he communi y is closely ela ed o being a co e de elope in mos cases. Two bene- i s men ioned by he in e iewees a e: (1) easie code inclusion and he eby a oid he need o ha - ing a p i a e code eposi o y, and (2) ecei ing mo e help om o he communi y membe s. D4 highligh s he impo ance o social posi ion in OSS communi y: “I hink i [ha ing a posi ion] helps a lo . I hink he e is a di e ence i , le s say, D2 asks o help, hen I will help him i I can. Bu i [De elope Name] om I ha e ne e eally hea d o , is asking o help hen my le el o e o is usually lowe . And pa o ha is because I know D2 pe sonally, and pa o ha is because I know ha he does a emendous amoun o wo k. My iew is ha i he needs help he dese es he help. And I hink i goes he o he way oo, i people a e mo e likely o help me because o he con ibu ions I ha e made and hey know ha I ha e been con ibu ing o a long ime. I hink i helps o ha e some so o s a us wi hin he communi y.” 5.5. S a egic ision The ole o s a egic ision on i m pa icipa ion is somewha ague in ou cases. Fi m’s s a egy could be how a i m de elops and deploy hei p oduc , i.e. how ex e nal esou ces a e used o educe de elopmen and main enance cos . The ision o i m’s s a egy needs o be aligned a no only manage ial bu also ope a ional le els. The ans e o s a egic isions is no clea ly e idenced in ou cases. Fo ins ance, a de elope D4 men ioned he spen signi ican o ice wo k hou s as well as spa e ime on con ibu ing o Wi esha k. He acknowledged he bene i s o he de elope s in his i m ecei ed om his pa ici- pa ion in he OSS p ojec and he ac ha he eely pa icipa ed in Wi esha k: “I is no an o icial pa o my job, bu a lo o he de elop- e s, es e s and he cus ome suppo people use Wi esha k ex ensi ely.” Howe e , his i m lacked o mal s a egies o decide how de elope s shall pa icipa e and de elop he OSS, wha code ha shall be con ibu ed back o o icial sou ces, and how o main ain he OSS knowledge base wi hin in he i m. 5.6. Ga ekeeping The pe cep ions o a ga ekeepe , who na iga es in o ma ion lows be ween his/he i m and ex e - nal ac o s, we e acknowledged by all in e iewees. While i ms migh ha e di e en needs and wo k p ac ices, ga ekeepe s a e he ones s ay in be- ween he i m and he OSS p ojec in some way, as shown in Figu e 9. D1 s a ed ha when his cowo ke s ound issues wi h he hi d pa y componen s, hey in o med D1, bu no p ojec manage s. D7 exp essed a simila pe cep ion: Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 53 “Yes, I ac as a b idge be ween [Company Name] and Samba and o wa d bugs/e o s o he com- muni y.” The ga ekeepe is o en an ac i e ac o in con ibu ing o he communi y, as men ioned by D2: “Many o ou co e de elope s a e wo king o smalle companies, and ha e a esponsibil- i y o he in e nal p o ocols ha hei company needs. (.. . ) I hink mos de elope s wo k indi id- ually, and ha e he ole o p o iding Wi esha k unc ionali y o he o he de elope s in he i m.” In a coope a i e manne , he ga ekeepe is he hub o in o ma ion and issues ha can be eached by di e en de elope s ac oss he o ganiza ions, as s a ed by D4: “Yes, e e ybody de ini ely knows ha I am he Wi esha k guy. All he de elope s, es e s and cus ome suppo people know ha hey can come o me i hey ha e Wi esha k issues (... )”. In i ms wi h mul iple de elope s ac i e in ups eam de elopmen , i.e. commi - ing o OSS p ojec s, he e is o en a ecognized ga ekeepe ole among hem. D5 men ioned: “In gene al when i comes o con ibu ing pa ches ups eam each de elope in [Company Name] is independen and can di ec ly app oach he ups eam p ojec ? The [Company Name] Samba package main aine usually has a ask o being he ga ekeepe o hose bugs ha ha e been epo ed agains [Company Name] p oduc s by he cus ome s o he suppo eams (. . . )” In his case, while code is con ibu ed independen ly by indi iduals in he i m, he bugs is managed by a ga ekeepe who submi s bug epo s on behal o he i m in o he OSS p ojec ’s bug acking sys em. In a compe i i e manne , ga ekeepe s would make su e ha no all p i a e sou ce code be e ealed o public. Fi ms migh con ibu e code ha ela e o co e componen s o OSS p oduc s, o u ili y unc ions. In a ypical scena io, i ms main ain hei p i a e eposi o ies, whe e many componen s a e pa s o i ms’ co e alues. Such componen s should no be e ealed, as men ioned by D4: “The majo i y o he s u I ha e w i en o Wi esha k has been pushed up? Bu you so o d aw a line in he s u ha is obscu e enough o no push. The only people who should be look- ing a ou p op ie a y p o ocol should be us?”. Some o he code is ega ded as p op ie a y and is e ained in he i m’s p i a e code eposi o y, due o echnical speci ic, o legal and au ho iza ion issues D2 men ioned: “Mainly p o ocol dissec o s o p o ocols used in ou equipmen , i he p o o- col is based on open p o ocol desc ip ions om 3GPP, ITU o IETF (RFC) i is conside ed OK o make an indi idual con ibu ion o OSS (.. . )”. Code which is no ele an , sensi i e o poo ly w i en would be il e ed ou by ga ekeepe s, as men ioned by D4: “The s u we do no send in is s u ha is no o in e es o anybody excep us (.. . ) And he o he pa is ha I do no hink he company would be h illed by a publica ion o hese p o ocols. In o de o push hose hings o Wi esha k I would need o ge au ho iza ion”. 5.7. Fi m awa eness Se e al in e iewees acknowledged he p esence o a leas ano he i m in he communi y (D1, D2, D3, D8, D9, D10). Howe e , de elope s e- ma k ha i is no he knowledge o wha o he i ms wo k o ha is aluable, a he i is he knowledge o wha business domain hey a e wo king wi hin. D2 eplied when was asked abou o he i m awa eness: “Yes, bu I do no know ha much abou he i ms o he o he de elope s. They ypically say ha hey wo k o Fi m X, and ha is i . Wha i m hey a e wo king o is no ha impo an o me.” D3 emphasized he po en ial alue o ha ing he i m awa eness: “(.. . ) I know ha D2 may ha e some ole as a con ac o Fi m X (...) I know ha D2 may be someone who is good a ge ing log iles o speci ic hings. In he pas when I was wo king wi h oice o e IP, I hough some imes he was able o gi e me some log iles om wi hin his company, bu I did no eally hink o him as he company ep esen a i e. I hink o him as a com- pany pe son who may be able o ge logs o me, like he does.” In Boo s ap, de elope s exp essed he conce n on how o he i ms we e doing ela ed o he web echnology, in o de o d aw lessons lea n o hei p oduc ision. D8 men ioned: “We ca e abou i o he company a e using his echnology in hei p oduc s, so we can lea n om hem (... ) We do no ca e i some guys jus wan o play wi h he echnology (. . . )” Ad- 54 Anh Nguyen-Duc e al. di ionally, he in e iewees we e asked i hey con- side ed ha hei con ibu ions could be used by o he i ms o gain compe i i e ad an age. The majo i y dismissed his pe cep ion, o example: “As Fi m X does no di ec ly con ol Wi esha k, I guess we ha e o be a bi ca e ul when we a e in con ac wi h o he de elope s (. . . ) I belie e, in he gene al case, ha you gain mo e om con- ibu ing o he de elopmen , ha e aining you code om he communi y”, s a ed by D2. A inal ema k by D5 abou he compe i i eness is: “Al- hough he e may be some compe i ion be ween companies, as enginee s we seek collabo a ion o mu ual bene i . We al eady know any ad- ancemen will be used by e e ybody, ha is no a p oblem, we ge back as much as we gi e ou .” 5.8. Collabo a ion Al hough collabo a ion wi hin an OSS commu- ni y is ypically in o mal and no planned, he e a e ma e s ha ha e o be decided upon. Fo ins ance, when he e is a new pos in a mailing lis , a de elope has o decide whe he o engage in he discussion wi h he o he s o no (essen- ially collabo a ing wi h hem). The awa eness o o he i ms in his aspec may p ospe he collabo a ion. Fi m- paid de elope s wi h simila needs and in e es s can collabo a e and d aw on each o he ’s abili ies. Knowing ha a de el- ope wo ks o a ce ain i m, and ha he can p o ide ce ain code a i ac s also in luences he collabo a ion. Es ablishing ela ionships o such aluable de elope s h ough collabo a ion is key. The e is a s ong desi e o e u n a o s and hono de elope ’s posi ions by assis ing hem when hey need help. Many comme cial i ms adop OSS, bu do no pa icipa e no con ibu e back o he OSS communi ies. Some o hese i ms collabo a e di- ec ly wi h o he s o de elop OSS-based p oduc s u he , wi h o wi hou pa icipa ing in he OSS communi y. How o pe o m he collabo a ion is an aspec i ms ha e o decide. As desc ibed abo e, he collabo a ion can ake place wi hin he OSS communi y using public o p i a e communi- ca ion channels, o ou side he communi y using p i a e channels and p i a e code eposi o ies. 5.9. Awa eness o compe i ion Fi ms wo king wi hin he same business domain a e o en compe i o s in he ma ke , and hus i is in e es ing o see how in luen ial he i m awa eness is when i ms come oge he in com- muni y based OSS p ojec s o de elop so wa e collec i ely. Su p isingly, i m-paid de elope s said ha hey pe cei ed o he de elope s as pa - ne s and/o iends a he han compe i o s. D5 poin ed ou ha he had me many de elope s a he OSS de elope con e ence, and conside ed many o hem as iends. D1 explained ha he did no make any dis inc ion be ween a i m-paid de elope and a olun ee ing de elope : “I hink o hem as de elope s, and no abou which i ms hey ep esen .” D7 said ha he would pe cei e o he s as pa ne s. D6 men ioned: “I ha e al- ways hough o o he s as pa ne s. E en mo e – I hink abou hem as colleagues.” D4, D8 and D9 sha ed simila hough s, and dismissed he pe cep ion o o he i m-paid de elope s as com- pe i o s: “I guess as hings ha e e ol ed we do ac ually compe e in some aspec s wi h some o hese people a his poin . Bu ha hasn’ e- ally occu ed o me much? I ha e no iced mo e people who end o be cus ome s o ou s, a he han ue compe i o s. We migh be compe i- o s wi hin some a eas, bu I ha e ne e eally hough abou i I guess”, s a ed by D9. The issue o compe i ion om a i m om somewhe e else in he wo ld migh no be sig- ni ican o a s a up and a SME who ocus on ha ing hei p oduc eleased as as as possible. Wi hou a clea ision on how hei ma ke o echnical ad an ages a e in luenced by sha ing and using OSS sou ce code, he conce n o com- pe i ion is no much ele an . D8 also men ioned: “You hink abou o he i ms as you compe i o s, bu I do no hink ha eally comes in o my in e ac ions eally. They ha e hei own use s somewhe e a ound he wo ld (...) I ha e some- imes seen con ibu ions om hei de elope s, bu I hink ha is good (... )”. Consequen ly, he coope i ion concep in hese OSS p ojec s migh be e y much coope a ion-dominan . Ano he obse a ion is ha he i m’s social posi ion is no used by any i ms o domina e Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 55 OSS de elopmen . D6 men ioned: “Be o e wo k- ing on Samba I used o hink ha big companies may ha e big in luence in OSS p ojec simply by ’buying’ co e de elope s. Now, ha I know mos o he people wo king on Samba, I know ha his is no easible.” Hence, ha ing a posi ion, o ’buying’ one, is no he way i ms ela e o no in luence he OSS de elopmen . 5.10. Consequen ac o s In e iewees acknowledged he bene i s o pa - icipa ing in OSS p ojec s, including knowledge sha ing, o ganiza ional lea ning and ask e ec- i eness. D2 men ioned ha many bes p ac ices ound in e iewing code and p ope commen s on commi s. He also app ecia ed he ac i eness le el o he p ojec wi h as eedbacks. The p ac ices a e acknowledged and b ough in o conside a- ion o imp o emen a his eam. Main aining an awa eness o he o he de elope s and wha hey a e cu en ly wo king on is also ecognized and is p omo ed by D6 in his i ms o a oiding duplica ed code ac oss he whole codebase. O - ganiza ional lea ning also occu s a he p ojec le el. When a i m obse es he pa icipa ion and in e ac ion o co e i ms in he OSS p ojec s, hey can in e s a egic ocus a eas om, i.e. ea u e eques s and applica ion cases. In ou cases, in-house p oduc de elopmen depends on he OSS p ojec s by (1) using ools as ou comes o he p ojec s o (2) in eg a ing and building hei p oduc s on op OSS com- ponen s. The dependence in e s ha a ask ha ela es o OSS codes is collec i e pe - o med and he ask scope is beyond he OSS p ojec . In a coope a i e-domina ed en i on- men , he ask will be done in an easie way. In a compe i i e-domina ed en i onmen , he awa eness o compe i o s migh be ha m ul o join ly comple ing he ask. Howe e , his is no di ec ly e iden om ou cases. 6. Discussions Table 5 summa izes ou indings in he compa i- son wi h exis ing li e a u e. While many indings con i m exis ing knowledge, hey also p o ide some no el indings. This sec ion will discuss ou indings based on ou opics: cen alized com- munica ion s uc u e in communi y-lead OSS p ojec s (Sec ion 6.1), modelling coope i ion in he con ex o OSS p ojec s (Sec ion 6.2), he ole o a ga ekeepe in implemen ing coope i ion s a egies (Sec ion 6.3) and i m con ibu ion s a egy in OSS p ojec s (Sec ion 6.4). Each sec- ion will discuss ou indings wi h ela ed wo k. The inal sec ion p esen s ou ac ions o add ess h ea s o alidi ies (Sec ion 6.5). 6.1. Cen alized communica ion s uc u e in communi y-lead OSS p ojec s Comme cial i ms pa icipa ing in commu- ni y-based OSS p ojec s collabo a e in a ious ways ac oss he o ganiza ional bounda ies. C ow- s on e al. s a ed ha communica ions s uc u e o a p ojec is an impo an elemen in unde - s anding p ojec ’s p ac ices [28]. In ou cases, he majo i y o he ac i i y in OSS p ojec s is gene a ed by a small subse o he i ms, and ha he emaining i ms pa icipa e wi h li le o none con ibu ion. Wi esha k and Samba demons a e a communica ions cen aliza ion s uc u e as in he onion-like social s uc u e model [28]. Oezbek e al. [60] in es iga ed ele en OSS p ojec s and e ealed ha he ole o a de elope in he co e laye migh be mo e impo an han he ac ha hey do (commi code, ix bug, answe emails, e c) mo e. Ou quan i a i e analysis o Wi esha k and Samba con i med hese esul s by showing he dominan con ibu ions o de elope s and i ms in he co e laye . Ou quali a i e da a e- ealed possible impo ance o hese de elope s in implemen ing i ms’ s a egies, i.e. collabo a ion o compe i ion. Dewan e al. [57] showed ha he he e ogenei y, which exis s be ween i m-paid de elope s and olun a y de elope s shapes he e olu ion o OSS communi y and i s p oduc . In ou case, we showed ha e en i m-paid de- elope s ha e signi ican con ibu ions o code commi s and communica ion, i is no signi i- can ly di e en be ween i m-paid and olun a y de elope s. F om communica ion s uc u e, his 56 Anh Nguyen-Duc e al. Table 5. Summa y o indings Findings Type Cu en knowledge OSS in as uc u e as ounda ion o bo h public and p i a e communica ion among i ms Con i ma ion S uc u es as hose in OSS enables he in e- g a ion o ex e nal esou ces [55]. Fi ms ac i i ies a e isible in OSS p ojec s Con i ma ion He e ogenei y exis s be ween i m-paid de- elope s and olun a y de elope s [56,57]. Some i ms in he co e posi ions, mos o i ms con ibu e li le New Onion-like s uc u e a de elope s le el [27,28]. Coope i ion exis s among i ms Con i ma ion S ong explici go e nance app oaches can di ec ly a ec o he i m’s bene i s [58]. Coope a ion-domina ed coope i ion among i ms a code and issue le els Con i ma ion Compe i ion o he same e enue model does no necessa y a ec collabo a ion wi hin OSS p ojec s [15–17,59]. Ga ekeepe s p o ide a mechanism o pe - o m coope i ion Con adic De elope s wi hin a i m need o be di ided o ake cha ge o ei he collabo a ion o com- pe i ion [19]. T us is he ounda ion o es ablishing com- munica ion, collabo a ion and also compe i- ion Con i ma ion T us as a success ac o s in collabo a ion in OSS p ojec s [38,42]. S a egic ision is no signi ican a de elop- e s’ le el New Sha ing s a egic ision is also c i ical o collabo a ion a eam le el. Fi ms gain social posi ion in OSS p ojec s, a oid me ging and bug ixes, impac on in- luencing de elopmen and ge suppo ed New Pe cei ed bene i is associa ed wi h bo h coope a i e and compe i i e a i udes [22, 40,41]. e eals a di e en inding om Dahlande ’s wo k [56]. 6.2. Modelling coope i ion in he con ex o OSS p ojec s Business li e a u e men ions he di icul y o iden- i ying coope i ion in a eal wo ld con ex [21]. Dagnino e al. [21] highligh ed ha coope i ion does no simply eme ge om joining compe i ion and collabo a ion, bu hey mix oge he o o m a new kind o s a egic in e dependence be ween i ms. We ag ee and illus a e o his iew by showing ha in OSS p ojec s, comme cial i ms ocus on ac i i ies ha c ea e a common alue wi h an awa eness o no sha ing hei echnical and legal sensi i e in o ma ion. F om ou cases, COSS alida es a he meso le el o s a egic collabo a ion, whe e i ms wi hin he same o simila domain collabo a e. Among an eceden ac o s om li e a u e, we highligh he ole o a s uc u al condi ion ia public and p i a e communica ion in as uc u es. The anspa en and e ec i e communica ion in as uc u e p o- ides a mechanism o coope i ion. Ou s udy desc ibes a compe i ion-domina ed ype o coope- i ion. E en when i ms a e awa e o hei com- pe i o s, he a i ude o collabo a ion is s ill o e - whelming. Valenca e al. aise a ques ion whe he i ms a e collabo a o s o compe i o s in so wa e ecosys ems [3]. A he equi emen enginee ing le el, he au ho s ound se e al signi ican chal- lenges among i ms wi hin he same collabo a i e ne wo k [3]. OSS p ojec s and i ms migh ha e di e gen in e es s bu i ms can manage o dis- co e a eas o con e gen in e es and be able o adap hei o ganizing p ac ices o collabo- a e [7]. In ou case, his is clea ly obse able a he ope a ional le el. The inding also ma ches wi h obse a ions by Linåke [15]. 6.3. The ole o a ga ekeepe in implemen ing coope i ion s a egies Beng sson e al. a gued ha indi iduals wi hin a i m could only ac in acco dance wi h one o he wo logics o in e ac ion a a ime, i.e., ei he o compe e o o collabo a e [19]. Ou obse a- Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 57 ion on a ga ekeepe ole gi es an al e na i e explana ion on how i ms manage such scena io. The i m’s s a egy can be lexible, o example ully open co e sou cing a one ime, and il e ing o sha ed code a ano he ime. The implemen a- ion o such s a egies is done ia he i m ga e- keepe , who does ac ual echnical con ibu ion o he communi y. The e o e, in con as wi h Beng sson’s indings, we ind ha i is possible o implemen a i m-le el dynamic in e ac ion ia in- di iduals in so wa e p ojec s. The ole o a ga e- keepe is discussed in he con ex o comme - cial dis ibu ed so wa e eams [61,62]. Ma czak e al. ound he ole o knowledge b oke s who would ha e a signi ican impac on in o ma ion low in equi emen -in e dependen eams [62]. In a con ex o i m- o- i m in e ac ion, we showed ha a ga ekeepe could na iga e he in o ma ion low beyond i m’s bounda ies. Nguyen-Duc e al. showed ou common asks o a ga ekeepe : ask nego ia ion, con lic esolu ion, ask- ela ed in- o ma ion na iga ion and bounda y objec se ups [61]. While he au ho s in es iga ed ga ekeepe s in a so wa e i m and a OSS p ojec sepa a ely, his wo k ocuses on bounda y spanning ac i i ies be ween he OSS communi ies and so wa e i ms. By in luencing he ga ekeepe s, managing code lows and in o ma ion lows, i ms can implemen compe ing o collabo a ing s a egies. 6.4. Fi m con ibu ion s a egy in OSS p ojec s The e exis some s udies cap u ing he phe- nomenon o comme cial i ms con ibu ing o OSS p ojec s. Linåke e al. in es iga ed con- ibu ion s a egies o i ms when pa icipa ing in OSS p ojec s [63]. The au ho s p oposed he Con ibu ion Accep ance P ocess (CAP) model o de e mine i sou ce code o any ypes o con- ibu ions can be con ibu ed o no . The CAP model bases on wo dimensions: (1) he bene- i s company can ecei e and (2) he knowledge behind he con ibu ions o acqui e and con ol [63]. While hese wo dimensions a e simila o ou model’s elemen s: pe cei ed bene i s (Sec ion 5.4) and ga ekeeping (Sec ion 5.6), ou model also explo e o he ac o s ha impac he ways i ms con ibu e o he OSS communi ies and col- labo a e wi h o he i ms. Muni e al. discussed how he openness o so wa e i ms migh help hem o gain bene i s om OSS communi ies om ou dimensions: (1) s a egy, (2) igge s, (3) ou comes, and (4) le el o openness. The model is simila wi h some elemen s in ou COSS model, i.e. s a egic ision, communica ion, ga e- keeping and consequen ac o s. Howe e , hese models do no cap u e he compe i ion s a egy ha i ms migh adop in OSS p ojec s. Unlike he p e ious wo k, ou COSS model p oposed a comp ehensi e iew on ac o s ha impac he s a egy o collabo a ion and compe i ion. 6.5. Th ea s o alidi y 6.5.1. Cons uc alidi y Th ea s o cons uc alidi y conside he e- la ionship be ween heo y and obse a ion, in case he measu ed a iables do no p o ide a good measu e o he ac ual ac o s [45]]. In a quali a i e s udy, cons uc alidi y can be hough o a “labeling” issue, as we migh ind he cons uc o he ou comes ha we belie e we a e ying o cap u e. A main as- sump ion in ou s udy lies in he way we iden- i y coope i ion among comme cial i ms. As he coope i ion concep comes om economic and business esea ch, we did no ha e a di- ec map om how he concep ope a ional- ize in SE esea ch. P e ious s udies ha men- ion e m “coope i ion” [3,15], do no p o ide he cons uc o his concep . Hence, o ou bes knowledge, his is he i s s udy in SE a emp o ope a ionalize his concep . We e- duced his isk by a de ail e iew and he iden- i ica ion o cha ac e is ics o coope i ion, he explo a ion o he con ex whe e he cons uc is in es iga ed. Bo h quan i a i e and quali a- i e da a was collec ed in concep ’s elemen s and summa ized in he end o desc ibe he model. We also include discussion wi h co-au- ho s and an expe in he en ep eneu ship in alida e ou obse a ion. The phenomenon is ope a ionalized based on public and p i a e communica ion among de- 58 Anh Nguyen-Duc e al. elope s pa icipa ed in OSS p ojec s. We we e awa e o o he communica ion channels, such as p i a e messaging, elephone and Skype, howe e , we do no ha e a easible way o quan i y his. We limi ed he in es iga ion in public collabo a ion whe e de elope s esponsed o he same mailing lis o commen on he same issue. Rega ding o he iden i ica ion o i m pa icipa ion, we used SNA wi h densi y me ics, such as deg ee cen- ali y and closeness [49]. O he ne wo k-based measu es o he same cons uc (e.g., ansi- i i y, compac ness, and connec edness) could be conside ed o enhancing he igo o his esea ch. We also used an unweigh app oach o pe o m SNA, which igno ed he i ms’ cha ac- e is ics, such as i m size, and business s a egy owa ds he OSS communi y. This could be con- side ed in u u e wo k, especially in i m-based OSS p ojec s. The isk o ope a ionaliza ion is educed by using a mixed me hod esea ch, including bo h quan i a i e and quali a i e da a. The in e ie- wees we e conduc ed wi h i ms om di e en social posi ion in OSS p ojec s, which inc ease he c edibili y in he obse a ion o phenomenon. The da a is limi ed a en in e iews. Howe e , we had eached da a sa u a ion [45] when in- e iewing Boo s ap case. Al hough, in e ie- wees we e selec ed om di e en ypes o OSS p ojec s, di e en company p o iles, we ound ha hei esponses we e consis en , which in- c ease ou con idence in he us wo hiness o he da a. 6.5.2. Ex e nal alidi y This h ea conside s he abili y o gene alize ou indings. The goal o his s udy is no o achie e s a is ical gene aliza ion, bu a he an analy ical gene aliza ion. This is pa icula ly im- po an when s udying a complex phenomenon, in ou case is coope i ion in OSS p ojec s. To a oid he bias on indings om a single case, we analyzed wo OSS p ojec s. Quali a i e da a was u he collec ed om he hi d OSS p ojec o imp o e he gene aliza ion. Wi h he in-dep h in es iga ion in bo h communi y and i ms’ sides o he p ojec s, we a e con iden abou he ex- plana ion powe o he COSS model o sim- ila con ex s. Ou OSS p ojec s p oduce a li- b a y, a amewo k and an applica ion, employ- ing GPL and MIT licenses. Ou cases ep esen o a communi y-ini ia ed OSS p ojec s, ha a e ini ia ed and lead by he communi y. Fu he esea ch should eplica e ou me hod on o he ypes o OSS p ojec s o explo e o he collabo- a ion and compe i ion scena ios. They a e also popula OSS p ojec s wi h yea s o ope a ion, hence he p oduc s and collabo a ion p ocess ha e been s able. The indings migh no be di- ec ly applicable o eme ging OSS p ojec s, o p ojec s ini ia ed by i ms. Resea ch on p ojec s wi h di e en ypes o OSS licenses migh lead o a a ie y in ou model. 6.5.3. Reliabili y This h ea conce ns abou he le el o which he ope a ional aspec s o he s udy, such as da a collec ion and analysis p ocedu es, a e epea able wi h he same esul s. The main da a collec ion was done as a pa o a mas e hesis. All in e - iews we e eco ded and ansc ibed e ba im in o de o make su e ha no da a educ ion occu ed p ema u ely. The ansc ip ion o he in e iews was e iewed and in e p e ed by he o he au ho . In case o ague s a emen s, one au ho is esponsible o ollow-up discussions wi h in e iewees o cla i ica ion. We used bo h quan i a i e da a abou communica ion among i ms in he p ojec and quali a i e da a om in e iews o i ms om di e en con ibu ion laye s. The da a iangula ion allows ou ind- ings ep esen he ue si ua ion o in es iga ed p ojec s. Mo eo e , he pape has gone h ough p oo - ead om se e al senio esea che s in he domain. Thei eedbacks help us o imp o e he pape signi ican ly since he i s d a . 7. Conclusions Coope i ion is an impo an opic in economics and business esea ch [19, 21], bu i is o e - looked in o he domains. In mode n so wa e indus y, he popula i y o de eloping so wa e Do So wa e Fi ms Collabo a e o Compe e? A Model o Coope i ion in Communi y-ini ia ed OSS P ojec s 59 p oduc s beyond i m’s bounda y makes coope i- ion a ele an heme. In his pape , we used bo h quali a i e and quan i a i e da a o in- es iga e coope i ion in OSS p ojec s. Fi s ly, we ound ha comme cial i ms pa icipa ing in communi y-ini ia ed OSS p ojec s collabo- a e in a ious ways ac oss he o ganiza ional bounda ies. While mos o i ms con ibu e li - le, a small numbe o i ms a e e y ac i e and accoun o la ge p opo ions o con ibu ion. I is also e iden ha i ms in e ac ac oss hei bounda ies in OSS p ojec s. Secondly, we p o- posed an empi ical model COSS o explain o oo causes o coope i ion in OSS p ojec s. The COSS model shows ha coope i ion is based on he i m awa eness, s uc u al condi ion o he OSS p ojec s and ope a ed by ga ekeepe s. The coope i ion is coope a ion-domina ed e en among i ms wo king in he same business do- main wi h simila business models. The indings ha e implica ions o esea ch. We o e a desc ip i e explana ion o how coope- i ion occu s and impac s in OSS p ojec s. We obse e ha so wa e i ms emphasize he co-c ea ion o common alue and pa ly eac o he po en ial compe i i eness in OSS p ojec s. The highligh o ou indings is he COSS model, which a gues ha compe i ion and collabo a ion can bo h be handled by ga ekeepe s. The ole o ga ekeepe s in c ossing o ganiza ional bound- a ies is s ill an in e es ing esea ch opic. Fo SE wi h abundan esea ch on OSS collabo a- ion and communica ion, he s udy on in e - i m coope i ion is a no el way o looking a he same da a sou ces and in as uc u es. The s udy also has implica ions o p ac i- ione s. We o e so wa e i ms insigh s abou di e en coope i ion s a egies obse ed in a com- muni y-d i en OSS p ojec . Fo ins ance, no all communica ion goes h ough he public channels in OSS p ojec s. Legal and secu i y sensi i e issues commonly go h ough p i a e o closed channels because o hei na u es. Fu he mo e, i ms should conside a ga ekeepe as an impo - an ole when hey plan o pa icipa e and gain bene i om OSS p ojec s. Fo u u e wo k, he nex s ep would be o alida e he COSS model wi h a la ge se o cases. Ou esea ch he e only uses h ee communi y-d i en OSS p ojec s, which limi s he gene aliza ion o indings. Mo eo e , a longi- udinal obse a ion on how coope i ion e ol es among i ms can p o ides knowledge ha goes beyond c oss-sec ional obse a ions. Las bu no leas , u he in es iga ion abou employing he ole o ga ekeepe s o coope i ion is needed o p o ide ac ionable guideline o success ul ope a- ion o in e - i m coope i ion. Fu u e wo k can also in es iga e OSS p ojec se ings ha a ec i m collabo a ion, i.e. OSS license, and ea u e eques mechanism. I would be in e es ed o see how hese ac o s could play a ole in ou model. Re e ences [1] D.G. Messe schmi and C. Szype ski, So - wa e Ecosys em: Unde s anding an Indispensable Technology and Indus y. Camb idge, MA, USA: MIT P ess, 2003. [2] S. Jansen, A. Finkels ein, and S. B inkkempe , “A sense o communi y: A esea ch agenda o so wa e ecosys ems,” in 31s In e na ional Con- e ence on So wa e Enginee ing – Companion Volume, 2009, pp. 187–190. [3] G. Valença, C. Al es, V. Heimann, S. Jansen, and S. B inkkempe , “Compe i ion and collabo- a ion in equi emen s enginee ing: A case s udy o an eme ging so wa e ecosys em,” in 22nd In e na ional Requi emen s Enginee ing Con e - ence (RE), 2014, pp. 384–393. [4] J.F. Lo aine Mo gann and P. Finnegan, “Explo - ing inne sou ce as a o m o in a-o ganisa ional open inno a ion,” in 19 h Eu opean Con e ence on In o ma ion Sys ems, 2011. [5] K. Manikas, “Re isi ing so wa e ecosys ems e- sea ch: A longi udinal li e a u e s udy,” Jou - nal o Sys ems and So wa e, Vol. 117, 2016, pp. 84–103. [6] H.H. Olsson and J. Bosch, “Ecosys em-d i en so wa e de elopmen : A case s udy on he eme ging challenges in in e -o ganiza ional R&D,” in So wa e Business. Towa ds Con inu- ous Value Deli e y, C. Lassenius and K. Smolan- de , Eds. Sp inge In e na ional Publishing, 2014, pp. 16–26. [7] S. O’Mahony and B.A. Bechky, “Bounda y o ga- niza ions: Enabling collabo a ion among unex- pec ed allies,” Adminis a i e Science Qua e ly, Vol. 53, No. 3, 2008, pp. 422–459. 60 Anh Nguyen-Duc e al. [8] B. And ea and R. C is ina, “Compa ing mo i- a ions o indi idual p og amme s and i ms o ake pa in he open sou ce mo emen : F om communi y o business,” Knowledge, Technology & Policy, Vol. 18, No. 4, Dec. 2006, pp. 40–64. [9] A.H. Ghapanchi, C. Wohlin, and A. Au um, “Re- sou ces con ibu ing o gaining compe i i e ad- an age o open sou ce so wa e p ojec s: An applica ion o esou ce-based heo y,” In e na- ional Jou nal o P ojec Managemen , Vol. 32, No. 1, 2014, pp. 139–152. [10] A.H. Ghapanchi, “Rallying compe encies in i - ual communi ies: A s udy o co e p ocesses and use in e es in open sou ce so wa e p ojec s,” In o ma ion and O ganiza ion, Vol. 23, No. 2, 2013, pp. 129–148. [11] R. Riehle, “The single- endo comme cial open cou se business model,” In o ma ion Sys ems and e-Business Managemen , Vol. 10, No. 1, Ma 2012, pp. 5–17. [12] A.H. Ghapanchi, C. Wohlin, and A. Au um, “Resou ces con ibu ing o gaining compe - i i e ad an age o open sou ce so wa e p ojec s: An applica ion o esou ce-based heo y,” In e na ional Jou nal o P ojec Man- agemen , Vol. 32, No. 1, 2014, pp. 139–152. [Online]. h p://www.sciencedi ec .com/science/ a icle/pii/S0263786313000380 [13] S. Ghobadi and J. D’Amb a, “Coope i i e e- la ionships in c oss- unc ional so wa e de el- opmen eams: How o model and measu e?” Jou nal o Sys ems and So wa e, Vol. 85, No. 5, May 2012, pp. 1096–1104. [14] K. Manikas and K.M. Hansen, “So wa e ecosys- ems? A sys ema ic li e a u e e iew,” Jou nal o Sys ems and So wa e, Vol. 86, No. 5, 2013, pp. 1294–1306. [Online]. h p://www.sciencedi ec . com/science/a icle/pii/S016412121200338X [15] L. Johan, R. Pa ick, R. Bjö n, and P. Mäde , “How i ms adap and in e ac in open sou ce ecosys ems: Analyzing s akeholde in luence and collabo a ion pa e ns,” in Requi emen s Engi- nee ing: Founda ion o So wa e Quali y. Cham: Sp inge In e na ional Publishing, 2016, pp. 63–81. [16] J. Teixei a, “Unde s anding collabo a ion in he open-sou ce a ena: The cases o WebKi and OpenS ack,” in P oceedings o he 18 h In e na- ional Con e ence on E alua ion and Assessmen in So wa e Enginee ing, 2014, pp. 52:1–52:5. [17] J. Teixei a, S. Mian, and U. Hy i, “Coope a ion among compe i o s in he open-sou ce a ena: The case o OpenS ack,” in In e na ional Con- e ence on In o ma ion Sys ems (ICIS), 2016. [18] A.M. B andenbu ge and B.J. Nalebu ., Co-ope i ion. New Yo k, USA: Doubleday, 1996. [19] M. Beng sson and S. Kock, “‘Coope i ion’ in business ne wo ks? To coope a e and compe e simul aneously,” Indus ial Ma ke ing Manage- men , Vol. 29, No. 5, 2000, pp. 411–426. [20] M. Zineldin, “Co-ope i ion: The o ganisa ion o he u u e,” Ma ke ing In elligence & Planning, Vol. 22, No. 7, 2004, pp. 780–790. [21] G.B. Dagnino and G. Padula, “Coope i ion s a - egy, a new kind o in e i m dynamics o alue c ea ion,” in The 2nd con e ence on Eu opean Academy o Managemen , 2002. [22] C.P. Lin, Y.J. Wang, Y.H. Tsai, and Y.F. Hsu, “Pe cei ed job e ec i eness in coope i ion: A su ey o i ual eams wi hin business o ganiza ions,” Compu e s in Human Beha - io , Vol. 26, No. 6, 2010, pp. 1598–1606. [Online]. h p://www.sciencedi ec .com/science/ a icle/pii/S0747563210001792 [23] W. Scacchi, “F ee/Open Sou ce so wa e de el- opmen : Recen esea ch esul s and eme ging oppo uni ies,” in The 6 h Join Mee ing on Eu- opean So wa e Enginee ing Con e ence and he ACM SIGSOFT Symposium on he Founda ions o So wa e Enginee ing: Companion Pape s, se . ESEC-FSE companion ’07. New Yo k, NY, USA: ACM, 2007, pp. 459–468. [24] J.Y. Moon and L. Sp oull, “Essence o dis ibu ed wo k: The case o he Linux ke nel,” Fi s Monday, No. 11, 2000. [Online]. h p://www. i s monday.dk/ojs/index. php/ m/a icle/ iew/801/710 [25] A. Mockus, R.T. Fielding, and J.D. He bsleb, “Two case s udies o Open Sou ce So wa e de- elopmen : Apache and Mozilla,” ACM T ansac- ions on So wa e Enginee ing and Me hodology, Vol. 11, No. 3, Jul. 2002, pp. 309–346. [26] G.K. Lee and R.E. Cole, “F om a i m-based o a communi y-based model o knowledge c ea ion: The case o he Linux ke nel de elopmen ,” O - ganiza ion Science, Vol. 14, No. 6, No . 2003, pp. 633–649. [27] J. Xu, Y. Gao, S. Ch is ley, and G. Madey, “A opological analysis o he Open Souce So wa e de elopmen communi y,” in P oceedings o he 38 h Annual Hawaii In e na ional Con e ence on Sys em Sciences, Jan. 2005, p. 198a. [28] K. C ows on and J. Howison, “The social s uc- u e o F ee and Open Sou ce So wa e de elop- men ,” Fi s Monday, Vol. 10, 2005. [29] L. Dabbish, C. S ua , J. Tsay, and J. He b- sleb, “Social coding in Gi Hub: T anspa ency and collabo a ion in an open so wa e eposi-