Two case studies of open source software development: Apache and Mozilla
Citation: Audris Mockus, Roy T. Fielding, James D. Herbsleb (2002/02/01) Two case studies of open source software development: Apache and Mozilla. ACM Transactions on Software Engineering and Methodology (TOSEM) (RSS)
DOI (original publisher): 10.1145/567793.567795
Semantic Scholar (metadata): 10.1145/567793.567795
Sci-Hub (fulltext): 10.1145/567793.567795
Internet Archive Scholar (search for fulltext): Two case studies of open source software development: Apache and Mozilla
Tagged: Computer Science (RSS)
The authors offer a series of empirical tests of a series of claims about the production dynamics in open source, using quantitative comparison of two large projects, Apache (web server) and Mozilla (browser). Using email archives and bug reports, they examine participation, team size, code ownership, productivity, defect density, and problem resolution intervals. One of the challenges the authors consider is how coordination can occur in the absence of the coordination mechanisms present in the industrial development model: system-level designs, defined processes, schedules, and so on. The authors observe that the level of coordination matches the material conditions of the participants: they are decentralized, thus have a decentralized decision-making pattern; their time availability varies, so the taking up of intensive and more regimented roles varies. The authors examined the contributions of Apache contributors versus many others and found that the top 15 developers contributed 80\% of the code in Apache, and likewise a similar core versus periphery emerged from commercial projects and that defect density rates for Apache before release were better than commercial projects but the rate after release (customer-facing) was worse; however the comparison is difficult due to differences in release cycle. The authors then examine Mozilla and test their refined hypotheses against this case, finding that Mozilla as a hybrid project has many more of the recognizable coordination mechanisms of industrial projects, but is also much larger; Mozilla developer productivity seems to exceed commercial productivity, a fact that may be attributable to engagement with a development community. Defect density in Mozilla was better than Apache, but Apache solved problems more quickly.
Theoretical and Practical Relevance
This article offers an empirical test of several debates about the role of open source models in shaping production dynamics. In addition to testing social theories, the article offers multiple clear measures and bases of comparison.