Abstract. Some free software and open source projects have been extremely successful in the past. The success of a project is often related to the number of developers it can attract: a larger community of developers (the `bazaar') identifies and corrects more software defects and adds more features via a peer-review process. In this paper two free software projects (Wine and Arla) are empirically explored in order to characterize their software lifecycle, development processes and communities. Both the projects show a phase where the number of active developers and the actual work performed on the system is constant, or does not grow: we argued that this phase corresponds to the one termed 'cathedral ' in the literature. One of the two projects (Wine) shows also a second phase: a sudden growing amount of developers corresponds to a similar growing output produced: we termed this as the `bazaar ' phase, and we also argued that this phase was not achieved for the other system. A further analysis revealed that the transition between `cathedral' and `bazaar ' was a phase by itself in Wine, achieved by creating a growing amount of new modules, which attracted new developers. 1Introduction Prominent free software (or open source software, OSS) projects such as Linux [32], Apache [27] and FreeBSD [18] have been extremely successful. Anecdotal evidence has been used in the past to characterize successful OSS projects: users/developers acting as "more eyeballs " in the correction of bugs, developers implementing new features independently, skillful project managers dealing with a mostly flat organization, and the resulting coordination costs [28]. Previous studies have provided empirical evidence on the process of successful OSS projects: the definition of various types of developers has been discussed for the Mozilla and the Apache projects, justifying different levels of effort [27], and claiming that the first type (core developers) contribute to the success of a system.

Abstract—We apply social network analysis (SNA) to examine the dynamics of leadership in distributed groups, specifically Free/Libre Open Source Software development projects, and its relation to group performance. Based on prior work on leadership in distributed groups, we identify leaders with those who make the highest level of contribution to the group and assess the degree of leadership by measuring centralization of communications. We compare the dynamics of leadership in two FLOSS projects, one more and one less effective. We find that in both projects, centralization was higher in developer-oriented communications venues than in user-oriented venues, suggesting higher degrees of leadership in developer venues. However, we do not find a consistent relation between centralization and effectiveness. We suggest that SNA can instead be useful for identifying interesting periods in the history of the project, e.g., periods where the leadership of the project is in transition. I.