Table of Contents
Acknowledgements………………………………………………………………………….xi
Chapter Synopsis……………………………………………………………………………xiii
CHAPTER 1: INCREASING AND DECREASING RETURNS TO SCALE………..1
INTRODUCTION………………………………………………………………………………….1
INCREASING RETURNS TO SCALE: THE ADVANTAGES OF BIGNESS …1
DECREASING RETURNS TO SCALE………………………………………………………6
Decreasing returns to scale due to coordination costs …………………………..6
Decreasing returns to scale due to reduced individual motivation ……….10
DOES PRODUCT MODULARITY MITIGATE THE ADVERSE EFFECTS OF
INCREASING SCALE?………………………………………………………………………….12
CHAPTER 2: LITERATURE REVIEW……………………………………………………….15
THE PRODUCTIVITY PARADOX IN SOFTWARE DEVELOPMENT …….15
MODULARITY IN ORGANISATION THEORY…………………………………….18
Product modularity and coordination costs………………………………………..22
Product modularity and productivity………………………………………………..27
Product modularity and group size……………………………………………………28
STUDYING MODULARITY IN FREE AND OPEN SOURCE SOFTWARE
DEVELOPMENT ………………………………………………………………………………..30
H1: Product modularity reduces coordination costs in FOSS projects……34
H2: Product modularity increases the potential number of contributors to
FOSS projects………………………………………………………………………………….37
H3: Product modularity has a positive effect on labour productivity in
FOSS projects………………………………………………………………………………….40
CONCLUDING REMARKS…………………………………………………………………..41
CHAPTER 3: RESEARCH METHODOLOGY……………………………………………..43
ANALYTICAL FRAMEWORK……………………………………………………………..43
Research Design………………………………………………………………………………43
Object of investigation……………………………………………………………………..44
Level of analysis………………………………………………………………………………46
WHY THE FREEBSD PROJECT?…………………………………………………………..47
MEASURING MODULARITY………………………………………………………………48
MEASURING COORDINATION COSTS……………………………………………….55
MEASURING DEVELOPERS GROUP SIZE…………………………………………..59
vii
MEASURING LABOUR PRODUCTIVITY……………………………………………..63
STATISTICAL ANALYSIS…………………………………………………………………….65
Sample selection………………………………………………………………………………66
Random-effects GLS regression…………………………………………………………67
Operationalisation……………………………………………………………………………69
CHAPTER 4: EMPIRICAL SETTING…………………………………………………………75
HISTORICAL BACKGROUND……………………………………………………………..75
ORGANISATIONAL STRUCTURE……………………………………………………….78
Core team……………………………………………………………………………………….79
Committers……………………………………………………………………………………..82
Outside contributors………………………………………………………………………..84
Ad hoc teams…………………………………………………………………………………..84
Hats………………………………………………………………………………………………..86
Maintainers…………………………………………………………………………………….86
TECHNICAL INFRASTRUCTURE………………………………………………………..87
Communication channels…………………………………………………………………87
Revision control ……………………………………………………………………………..88
Reporting & managing defects………………………………………………………….88
Testing……………………………………………………………………………………………88
Distribution channels………………………………………………………………………89
DEVELOPMENT PROCESS …………………………………………………………………89
SCALE ……………………………………………………………………………………………….91
CHAPTER 5: MODULARITY AND COORDINATION COSTS IN FREEBSD. .95
INTRODUCTION………………………………………………………………………………..95
QUALITATIVE ANALYSIS…………………………………………………………………..99
CONCLUDING REMARKS…………………………………………………………………104
CHAPTER 6: MODULARITY AND GROUP SIZE IN FREEBSD…………………107
INTRODUCTION………………………………………………………………………………107
QUALITATIVE ANALYSIS…………………………………………………………………111
QUANTITATIVE ANALYSIS………………………………………………………………113
A SUMMING UP……………………………………………………………………………….127
REVERSING THE TERMS OF THE PROPOSITION……………………………..130
Scale considerations……………………………………………………………………….134
CONCLUDING REMARKS…………………………………………………………………137
CHAPTER 7: MODULARITY AND LABOUR PRODUCTIVITY IN FREEBSD
viii
……………………………………………………………………………………………………………..139
SETTING OF THE PROBLEM…………………………………………………………….139
QUALITATIVE ANALYSIS…………………………………………………………………142
QUANTITATIVE ANALYSIS………………………………………………………………149
Scale considerations……………………………………………………………………….150
EFFECT OF MODULARITY ON CORE DEVELOPERS PERFORMANCE154
Scale considerations……………………………………………………………………….156
CONCLUDING REMARKS…………………………………………………………………160
CHAPTER 8: DOES BROOKS’ LAW HOLD IN FREEBSD?………………………..161
INTRODUCTION………………………………………………………………………………161
DISAGGREGATING CORE DEVELOPERS’ PRODUCTIVITY ……………..164
EFFECT OF GROUP SIZE ON CORE DEVELOPERS PERFORMANCE….167
Scale considerations……………………………………………………………………….169
Does modularity negate Brooks’ Law?……………………………………………..173
CONCLUDING REMARKS…………………………………………………………………174
CHAPTER 9: THE EMERGENCE OF GOVERNANCE……………………………….175
INTRODUCTION………………………………………………………………………………175
INFORMAL GOVERNANCE PHASE (1993-2000)………………………………..176
DEMOCRATIC GOVERNANCE PHASE (2000-TO DATE)……………………178
THE IMPERATIVE OF AUTONOMY………………………………………………….185
AUTHORITY AND LEGITIMACY………………………………………………………190
CONCLUDING REMARKS…………………………………………………………………200
CHAPTER 10: CONCLUSIONS………………………………………………………………..203
SUMMARY REVIEW OF RESULTS ……………………………………………………203
EFFECT OF PRODUCT STRUCTURE ON GROUP DYNAMICS……………205
Decentralisation made scalable………………………………………………………..205
Modularity reinforces the emergent division of labour………………………207
Effect of product modularity on labour productivity…………………………208
EFFECT OF GROUP DYNAMICS ON PRODUCT STRUCTURE……………209
Product structure mirrors organisational structure……………………………209
Product structure as coordination mechanism………………………………….211
Why not in small-scale development conditions?……………………………..213
EFFECT OF GROUP SIZE ON LABOUR PRODUCTIVITY……………………214
Brooks’ Law revisited……………………………………………………………………..214
GENERALISABILITY…………………………………………………………………………223
ix
Across community of FOSS projects………………………………………………..223
Beyond the realm of FOSS………………………………………………………………224
EPILOGUE…………………………………………………………………………………………….227
SUMMARY……………………………………………………………………………………………239
SAMENVATTING (SUMMARY IN DUTCH)……………………………………………245
APPENDICES…………………………………………………………………………………………251
APPENDIX I: THE FREEBSD LICENSE……………………………………………….253
The FreeBSD Copyright………………………………………………………………….253
APPENDIX II: RELEASE RATE (1993-2003)………………………………………..255
APPENDIX III: COMMITTERS ADDED AND REMOVED PER MONTH
(2000-2003)……………………………………………………………………………………….257
APPENDIX IV: CORE DEVELOPERS SURVEY……………………………………259
Email Questionnaire………………………………………………………………………259
Analysis………………………………………………………………………………………..260
Results………………………………………………………………………………………….260
Collected replies…………………………………………………………………………….261
APPENDIX V: BIBLIOGRAPHICAL REFERENCES……………………………..267
CURRICULUM VITAE……………………………………………………………………….299
Abstract
My dissertation looks at the Governance Structures of Free/Open Source (FOSS) Development, based on a case study of FreeBSD, a large FOSS project. More specifically, it examines 3 well-known theories. The 1st theory [decreasing returns to scale] holds that increasing the number of persons working together results in a productivity drop due to (a) increased coordination costs and (b) reduced individual motivation. The 2nd theory [iron law of oligarchy] holds that a group’s ability to self-organise diminishes as it gets larger, thus necessitating hierarchical coordination. And the 3rd theory [modularity theory] holds that modularity (a design principle implemented by breaking down a product into autonomous components) increases the potential number of persons working on a distributed project and has a positive effect on their labour productivity, because it allows them to work independently of each other, with little or no need for active coordination. To test these theories, we looked at the development of the FreeBSD operating system project over a period of 15 years, in which time the number of persons developing it increased dramatically from about a dozen in 1993 to several hundred in 2010. What we found, with respect to the 1st theory, is that indeed the increase of FreeBSD developers resulted in a fall in group performance but the cause of this was the disproportionate increase of ‘lower-contribution’ participants over time, and not increased coordination costs or reduced individual motivation. With respect to the 2nd theory, we ascertained that the increase of scale of the FreeBSD group in numbers did not result in hierarchy, which outcome is accounted for by the normative standard of individual autonomy of action and the fact that the distributed environment in which FOSS projects operate does not foster hierarchy. With respect to the 3rd theory, we found that indeed modularity increases the potential number of persons working on a project and has a positive effect on their performance, as it allows them to work independently of each other, with little or no need for active coordination.
Roadmap onderwerpen
Veerkrachtig ontwerp en onderhoud
