Crossover from Endogenous to Exogenous Activity in Open-Source Software Development

Abstract

We have investigated the origin of fluctuations in the aggregated behaviour of an open-source software community. In a recent series of papers, de Menezes and co-workers have shown how to separate internal dynamics from external fluctuations by capturing the simultaneous activity of many system's components. In spite of software development being a planned activity, the analysis of fluctuations reveals how external driving forces can be only observed at weekly and higher time scales. Hourly and higher change frequencies mostly relate to internal maintenance activities. There is a crossover from endogenous to exogenous activity depending on the average number of file changes. This new evidence suggests that software development is a non-homogeneous design activity where stronger efforts focus in a few project files. The crossover can be explained with a Langevin equation associated to the cascading process, where changes to any file trigger additional changes to its neighbours in the software network. In addition, analysis of fluctuations enables us to detect whether a software system can be decomposed in several subsystems with different development dynamics.