This research is about code production in top-level open source communities of The Apache Software Foundation (ASF). We extensively analyzed Subversion repository logs from 70 top-level Apache open source projects in the ASF from 2004 to 2009. Based on interactions in code production during one-year periods we constructed networks of file co-authorship that gave us access to the organization of Apache open source communities. This allowed us to measure graph level properties, like hier- archy and clustering, and their influence on the outputs of code production.
Apache communities are groups of individuals that organize their code production efforts in order to develop enterprise-grade open source software. The ASF explains the success of its communities and the software they produce by claiming to have instituted a meritocracy that brings contributors together in a way that significantly influences code production, namely by building communities instead of only focusing on technical properties of the source code like modularity. Self-organization theory has found that the role of institutions is minor. In this research we test and confirm the theory of self-organization, and find that the meritocracy institution does not influence code production.