Latent Social Structure in Open Source Projects

TitleLatent Social Structure in Open Source Projects
Publication TypeConference Paper
Year of Publication2008
AuthorsBird, C, Pattison, D, D'Souza, R, Filkov, V, Devanbu, P
Secondary TitleSIGSOFT '08/FSE-16: Proceedings of the 16th ACM SIGSOFT Symposium on Foundations of Software Engineering
Pagination24–35
PublisherACM
Abstract

Commercial software project managers design project organizational structure carefully, mindful of available skills, division of labour, geographical boundaries, etc. These organizational “cathedrals” are to be contrasted with the "bazaar-like" nature of Open Source Software (OSS) Projects, which have no pre-designed organizational structure. Any structure that exists is dynamic, self-organizing, latent, and usually not explicitly stated. Still, in large, complex, successful, OSS projects, we do expect that subcommunities will form spontaneously within the developer teams. Studying these subcommunities, and their behavior can shed light on how successful OSS projects self-organize. This phenomenon could well hold important lessons for how commercial software teams might be organized. Building on known well-established techniques for detecting community structure in complex networks, we extract and study latent subcommunities from the email social network of several projects: Apache HTTPD, Python, PostgresSQL, Perl, and Apache ANT. We then validate them with software development activity history. Our results show that subcommunities do indeed spontaneously arise within these projects as the projects evolve. These subcommunities manifest most strongly in technical discussions, and are significantly connected with collaboration behaviour.

Notes

We first identified the projects of interest and mined the developer mailing list archives and source code repositories of each of the projects. Next, we filtered the mailing list messages and created a social network of the participants over 3-month intervals. We then calculated the community structure of each social network. Following that, the relevance of the divisions of participants was evaluated quantitatively using mined source code development data and qualitatively by manual methods.

Full Text
AttachmentSize
PDF icon bird2008lss.pdf474.25 KB