A longitudinal analysis of the dependency concentration in smaller modules for open-source software products

Abstract

Our recent studies on single releases of multiple open-source software (OSS) products showed a higher concentration of dependencies in smaller modules. For one of the products, it was observed that an isolatable and observable refactoring initiative exacerbated this concentration inequality. In this paper, we empirically investigate the dependency concentration in smaller modules from a longitudinal perspective: (1) whether this concentration inequality always exists over product life time; (2) how it changes. We hypothesize that the concentration inequality should either remain at same levels or increase over time. This is because large-scale and long-lived software products usually go through some degree of continuous and intermittent refactoring. Our results show that dependencies concentrate in smaller classes in all releases, and this concentration inequality generally increases over successive releases. We suggest that software practitioners continuously pay a higher QA attention to smaller modules. We also recommend increasing such QA focus as a product matures and goes through refactoring activities.