Managing the Boundary of an 'Open' Project

Siobhán O’Mahony and Fabrizio Ferraro have written an excellent paper on evolution of the Debian project. In Beyond Majority Rule Sheeran notes that the growth and control of decision making among Quakers, who otherwise were individualistic and loosely coupled, arose from the threat of government persecution. O’Mahony and Ferraro note that the need for managing boundaries in an “open” technical project arise in order to protect the project from trojans: folks who join the community for malicious purposes such as introducing back-doors.

We examine the project’s face-to-face social network during a five-year period (1997-2001) to see how changes in the social structure affect the evolution of membership mechanisms and the determination of gatekeepers. While the amount and importance of a contributor’s work increases the probability that a contributor will become a gatekeeper, those more central in the social network are more likely to become gatekeepers and influence the membership process.(2)

They also consider social connectedness by looking at the Debian Developer PGP keyring:

Betweenness centrality is a measure that synthetically captures the structural position of developers in the social network and each individual’s ability to potentially broker information and exert social influence. In this context, betweenness centrality is a measure of an individual’s ability to link disconnected parts of the network through face-to-face interaction3. Betweenness centrality (Freeman, 1979; Marsden, 1982; Wasserman and Faust, 1994) measures the extent to which an actor can broker communication between other actors. (14)

There research shows that one’s “betweenness centrality” and the popularity of one’s packages is predictive of gatekeeper status, though oddly enough experience in the community is not:

The popularity of one’s package is also predictive of NMC status. For every 100 people who use a developer’s package, he or she is 4% more likely to become a NMT member (3% in 2003). Tenure likely had a negative effect because those who joined the project more recently were more likely to be aware of the problems with admitting new members. In 2002, these results are confirmed, even though the magnitude of the effect of centrality is smaller (Odds ratio=1.47). (27)

What’s nice about this paper is it includes a theoretical treatment (open science), historical exposition (evolution of Debian organization), and social network analysis augmented with ethnography (interviews). This is what I also liked about Stephen Lansing and John H. Miller’s Cooperation in Balinese Rice Farming: a game theoretical model of upstream/downstream rice farmers given the variables of water deprivation and pest damage validated by field interviews.

Comments !