NETWORK ANALYSIS OF THE SOURCEFORGE COMMUNITY
By Chris Zachor
Overview Introduction Background
Open Source Software The SourceForge community and network
Previous Work What can be done different? Related Work Conclusion Questions
Introduction Goal: Use network analysis to better
understand the SourceForge community developers
Identify key developers and groups of developers who create popular open source applications
Open Source Software Open Source (OS) Software continues to
be a popular alternative to standard commercial software
Many OS alternatives to traditional closed source projects exist
osalt.com provides a convenient database for this
The SourceForge Community
A website to help promote collaboration between developers of OS projects
A repository for OS projects Developers: revision control, bug
tracking, donation system, etc. Users: bug reporting, recommendations,
commenting, etc.
The SourceForge Network Multiple networks can be formed from the
SourceForge community• Project-Developer network• Developer network• Project network• Lots of interesting data to be collected
from the website such as total downloads of a project, length of developer membership, recommendations, etc.
Project-Developer Network A bipartite graph with two groups of
vertices: projects and developers An edge indicates the developer works on
that project
Developer Network A collaboration network Edges are formed where one developer 1
has worked with developer 2
Project Network An edge can represent a related project An edge can represent projects that share
a developer Or perhaps an edge can represent a
related project
Previous Work on SourceForge The open source group at Notre Dame Used network analysis as a tool to
understand the Open Source Software phenomenon and predict growth over time
Monthly data dumps directly from SourceForge.net
What Can Be Done Different?
The latest paper produced concerning network analysis was in 2007
The project count has more than doubled in size to ~250,000 projects (from ~90,000 in 2007).
What Can Be Done Different?
Their main concern was with how the network was evolving
Focus was on the change in measures from month to month
No interpretation of data
Related Work M. E. J. Newman Scientific Collaboration Networks 4 Major Databases spanning 5 years Collaboration network using authors who
have worked together on a single paper Explored what fields were producing
more papers, what fields collaborated more, etc.
Related Work Obermeier et al. University College Dublin Co-authorship between departments at UCD They wanted to understand the
interdisciplinary publication culture within the University
Looked at brokerage individuals and how they play a part in their own departments
Found these brokerage individuals to be most central within their own departments
Related Work Gao and Madey Network analysis of SourceForge Used as a tool to understand the open
source movement Documented the growth of the
SourceForge community Structural analysis, centrality analysis,
path analysis They did not interpret the data
Related Work Xu, Christley, and Mady Network analysis of the SourceForge
community Attempt to explain the success and
efficiency of OS development practices Noted that the SourceForge Network is a
scale free network Also noted the presence of the small
world phenomenon within the community
Related Work Xu, Christy, and Madey continued Observed that co-developers and active
users were a major factor in large scale projects
Meanwhile, project leads and core developers were largely involved in small projects
Conclusion While previous studies were focused on
growth and why the process is a success, this study will focus on how key developers and groups play a part in creating popular software
Many attributes not looked at in previous studies
Questions?
Anyone?