Impact of Release Intervals on Empirical Research into Software Evolution, with Application to the Maintainability of Linux.

IET Software, 3(1):58-66, February 2009.

Larry G. Thomas, Stephen R. Schach, Gillian Z. Heller, Jeff Offutt


In most empirical research on software evolution, analysis of the data is performed with respect to the release sequence number, rather than the release date. This distinction is important when the intervals between release dates vary widely, as is generally the case with open-source software. A widely cited paper on the maintainability of Linux was published in this journal in 2002. The paper showed that, whereas the size of the Linux kernel grew linearly with respect to release sequence number, the amount of common coupling grew exponentially. In view of the adverse effect of common coupling on maintainability, the conclusion drawn in that paper was that Linux needed to be refactored with minimal common coupling. Here we show that, if the same data is analysed with respect to release date, the amount of common coupling grows linearly; hence, there is no need to refactor Linux to promote maintainability. We also analyse three stable series of Linux releases, and observe that the size and the common coupling grow linearly. We conclude that rates of growth should be computed with respect to temporal variables, such as release date.

Back to my home page.