Tracing distributed collaborative development in apache software foundation projects

Developing and maintaining large software systems typically requires that developers collaborate on many tasks. During such collaborations, when multiple people work on the same chunk of code at the same time, they communicate with each other and employ safeguards in various ways. Recent studies have considered group co-development in OSS projects and found that it is an essential part of many projects. However, those studies were limited to groups of size two, i.e., pairs of developers. Here we go further and characterize co-development in larger groups. We develop an effective methodology for capturing distributed collaboration beyond groups of size two, based on synchronized commit activities among multiple developers, and apply it to data from 26 OSS projects from the Apache Software Foundation. We find that distributed collaborations is prevalent, but not as frequent as expected. We also find that while in distributed collaborative groups, developers' behavior is different than when programming alone, e.g., high developer focus on specific code packages associates with lower team participation, while packages with higher ownership get less attention from groups than from individuals. Finally, we show that productivity effort during co-development is more often lower for developers while they co-develop in groups. To verify our results we use both quantitative and qualitative methods, including a developer survey. We conclude that these methods and results can be used to understand the effects of the collaborative dynamic in OSS teams on the software engineering process. Our code, along with our datasets and survey is available at http://www.gharehyazie.com/supplementary/teamwork/.