Evolutionary Trends in the Collaborative Review Process of a Large Software System

In this paper, we study the evolutionary trends in the collaborative review process of a large open source software system. As expected, the number of reviews, the number of reviews commented on, as well as the number of reviewers, and the interactions between them show increasing trends over time. But unexpectedly, levels of clustering between developers in their interaction networks show a decreasing trend, even as connections between them increase. In the context of our study, clustering is an indicator of developer collaboration, whereas connection points to how intensely developers work together. Thus the trends we observe can inform how developer interactions become concentrated around specific units of work as the project progresses. The dichotomy between the simultaneous increase in connection and decrease in clustering also points to the interplay between collective and individual efforts in the review process, and the distinct nature of peer review in the software development life cycle.