Discriminating Development Activities in Versioning Systems: A Case Study

When characterizing the coding activity of developers on a software project, a common technique is to study the transactions in the versioning system. However, not all transactions are equal, neither represent the same kind (both in quality and quantity) of activity. In this paper, a methodology for classifying the interactions of developers with versioning systems is proposed, based on the analysis of the textual descriptions (comments) attached to each transaction. Some results of applying this methodology to a specific case (the FreeBSD CVS repository) are presented, alongside with some ideas about how to use this technique for improving effort estimation and other research lines.