What's wrong with git?: a conceptual design

Abstract

It is commonly asserted that the success of a software development project, and the usability of the final product, depend on the quality of the concepts that underlie its design. Yet this hypothesis has not been systematically explored by researchers, and conceptual design has not played the central role in the research and teaching of software engineering that one might expect. As part of a new research project to explore conceptual design, we are engaging in a series of case studies. This paper reports on the early stages of our first study, on the Git version control system. Despite its widespread adoption, Git puzzles even experienced developers and is not regarded as easy to use. In an attempt to understand the root causes of its complexity, we analyze its conceptual model and identify some undesirable properties; we then propose a reworking of the conceptual model that forms the basis of (the first version of) Gitless, an ongoing effort to redesign Git and experiment with the effects of conceptual simplifications.