Design Evolution of an Open Source Project Using an Improved Modularity Metric

Modularity of an open source software code base has been associated with community growth, incentives for voluntary contribution, and a reduction in free riding. As a theoretical construct, it links open source software to other domains of research, including organization theory, the economics of industry structure, and new product development; however, measuring the modularity of an open source software design has proven difficult, especially for large and complex systems. Building on previous work on Design Structure Matrices (DSMs), this paper describes two contributions towards a method for examining the evolving modularity of large-scale software systems: (1) an algorithm and new modularity metric for comparing code bases of different size; and (2) evolution analysis of Apache Tomcat to illustrate the insights gained from this approach. Over a ten-year period, the modularity of Tomcat continually increased, except in three instances: with each major change to the architecture or implementation, modularity first declined, then increased in the subsequent version to fully compensate for the decline.