Yongqin Gao and Greg Madey: “Network Analysis of the SourceForge.net Community”

Gao and Madey use standard network analysis techniques to analyze characteristics of the SourceForge.net community, which serves as the primary collaboration network for many open source projects. They use Structure and Centrality analysis to analyze the structure of the network itself, irrespective of time (i.e., utilizing a snapshot of network topography; their data is from 2006), and Path analysis to trace the change in certain key measures over time, in order to discover trends.

The authors note that two of the most important measures for analyzing collaborative communities are the Diameter of the network and its Clustering Coefficient. The diameter, or average length of the shortest paths between any pair of nodes, is an effective measure of network communication ability—specifically “efficiency of information propagation.” The Clustering Coefficient is the ratio of the number of links to the total possible number of links among nodes close to each other. The higher the Clustering Coefficient, “the more connected the network is.” (197)

The results of the analysis demonstrated that the Sourceforge.net network maintained a relatively small diameter (a sign of high efficiency of information propagation) and a very high degree of clustering. Generally, as networks grow in size, the diameter increases as well; however, path analysis indicates that the diameter is actually decreasing even as the network grows rapidly. This suggests an evolution toward greater efficiency: the time it takes for one developer to spread an idea to other developers is decreasing. At the same time, SourceForge’s clustering coefficient is increasing, indicating that the network is becoming increasingly connected. Finally, the authors found that average project size and average number of projects a single developer participated in are decreasing, possibly providing a clue as to the causes of network efficiency gains. Open Source development, inherently based upon collaborative networking, appears to scale very well: as network size and complexity increase, connectivity and efficiency increase, contrary to normal network effects.


Tagged with:

Leave a Reply

Comments are closed.