Using our expert knowledge of Linux development, we present additional interpretations of the results.
Using simple, established techniques like measuring the amount of contributed code to evaluate developer influence turned out to result in incorrect conclusions. By applying PageRank analysis to developer networks we are able to produce results that more closely reflect the real-world perception of influential developers. The tables shown on the right present the results for Linux version 3.7. The page rank analysis applied to the developer network is able to correctly identify a number of highly influential developers, indicated by their known roles in the community. In contrast, ranking developers according to commit count fails to identify many of the known influential developers. In summary, the rich information embedded in a developer network's structure is able to capture an authentic representation of developer influence that is not captured by more common approaches such as commit count.
Rank | Developer | Tag Count | Role |
---|---|---|---|
1 | Linus Torvalds | 974 | Central Maintainer |
2 | Greg KH | 3252 | Linux Foundation Fellow |
3 | Mauro Carvalho Chehab | 1624 | Tooling Contributor |
4 | David S. Miller | 1210 | Networking Core Maintainer |
5 | Herbert Xu | 162 | Networking Core Contributor |
6 | Antti Palosaari | 528 | Individual Contributor |
7 | Daniel Vetter | 558 | Intel DRM Drivers Maintainer |
8 | Andrew Morton | 902 | General Coordinator |
9 | Felipe Balbi | 588 | Maintainer for various USB aspects |
10 | Samuel Ortiz | 282 | IRDA, MFD and NFC co-Maintainer |
Rank | Developer | Commit Count |
---|---|---|
1 | Hartley Sweeten | 471 |
2 | Antti Palosaari | 219 |
3 | Al Viro | 179 |
4 | Wei Yongjun | 147 |
5 | Sachin Kamat | 142 |
6 | Mark Brown | 136 |
7 | Eric W. Biederman | 130 |
8 | David Howells | 126 |
9 | Hans Verkuil | 123 |
10 | Daniel Vetter | 123 |
Using our approach, we were able to correctly infer from the network structure, a number of communities that have shared development responsibilities. The communities that are largely responsible for the KVM hypervisor, Ext4 filesystem, S390 mainframe architecture support, and staging device drivers are shown on the right. This result supports the notion that our approach is able to capture real-world collaboration and identify communities of people with shared responsibility.