Hierarchy and centralization in free and open source software team communications

From AcaWiki
Jump to: navigation, search

Citation: Kevin Crowston, James Howison (12/2006) Hierarchy and centralization in free and open source software team communications. Knowledge, Technology & Policy (RSS)
DOI (original publisher): 10.1007/s12130-006-1004-8
Semantic Scholar (metadata): 10.1007/s12130-006-1004-8
Sci-Hub (fulltext): 10.1007/s12130-006-1004-8
Internet Archive Scholar (search for fulltext): Hierarchy and centralization in free and open source software team communications
Download: http://link.springer.com/10.1007/s12130-006-1004-8
Tagged: Computer Science (RSS)

Summary

This article describes a series of predictions and assumptions made both by practitioners and empirical researchers about the structure and communication dynamics of open source projects. Using a multi-site, multi-project dataset, the authors use network measures to test whether or not open source projects tend to be hierarchical, centralized and display mutuality of communication. They find that in general multiple structures are observed, but larger projects tend to be modularized i.e. less centralized (although it's unclear whether modularity tends to precede or follow growth). Code-level participation was observed to be centralized with a small subset of participants developing the bulk of the code, and that communications between contributors and bug-reporters tended not to be mutual, suggesting hierarchy. A measure of 'lubness' (least-upper-boundedness), interpreted to mean the degree to which a single person tended to eventually participate in each discussion, 

Theoretical and Practical Relevance

This article offers direct answers to several questions about how open source projects are structured and succeed, serving to debunk some assumptions about them as free-wheeling to the point of being chaotic; instead we see a diverse range of structures with some regularity observed in participation patterns consistent with the long tail: many people contributing only a little (in the form of bug reports) and a few people contributing quite a lot (in the form of the bulk of the code).