Final Thesis: Collaboration Networks in Open Source Projects’ Issue Trackers

Abstract: Open source software nowadays is increasingly used outside its own ecosystem. It is used by governmental departments and companies, for commercial projects and as a part of critical infrastructure, like encryption libraries. This usually makes it necessary to increase the collaboration with open source communities but it is also required to asses new risks, which arise from using a software, maintained mainly by volunteers. In order to achieve this we need a better understanding of open source communities and about how they organize and structure themselves, whether they depend on key personality and whether they form subcommunities? How do such communities change their behavior over time, during growth and which is the most efficient form of structuring to handle reported issues? We collect issue tracker data and use it to create sequences of social networks for 6007 projects present on Github.com and for 120 projects from the Apache Software Foundation. Based on metrics to quantify the strength of subcommunities and centralization we study the general structure of open source communities and how they might correlate, but also their behavior over time. We compare the communities based on an efficiency metric to gain information about preferable structures. Our results show that most open source communities avoid to organize themselves in subcommunities while they are highly dependent on a few key personalities. The results of both metrics do indeed correlate, which means that if a community has strongly distinct groups it is unlikely to be highly centralized. But neither  the few in subcommunities organized projects nor any other organizational type show a significantly higher efficiency. Although projects grow over time, they show only little internal structural change. The stability of open source communities and the strong avoidance of subcommunities are unexpected results, since it highly contradicts recent related studies and therefore requires more research to better predict the development of projects. For the assessment of projects the dependency on strong key members for the stability may prove helpful when it comes to indicating and analyzing major changes.

Keywords: Open Source, Issue Tracker, Bug Tracker, Social Network Analysis

PDFs: Master Thesis, Work Description

Reference: Björn Meier. Collaboration Networks in Open Source Projects’ Issue Trackers. Master Thesis, Friedrich-Alexander-Universität Erlangen-Nürnberg: 2015.