Palantir: Enhancing configuration management systems with workspace awareness to detect and resolve emerging conflicts

Software development requires multiple developers to work in concert on highly interdependent tasks and artifacts, which makes coordination a crucial part of any software development process. Coordination breakdowns are responsible for a wide variety of software conflicts that, in turn, lead to cost overruns and project delays. This dissertation focuses on those software conflicts that arise because of parallel development during the implementation phase. More specifically, it focuses on direct conflicts (when the same artifact is concurrently modified by multiple developers) and indirect conflicts (changes in one artifact negatively affecting concurrent changes in another artifact) as they arise when developers work together using a configuration management system.; Our work is based on the insight that the use of current configuration management systems has the undesired side effect of isolating developers in their private workspaces. As a result, they remain unaware of emerging conflicts until changes are complete and have to be brought together, build, and tested. Conflicts are therefore detected much later in time. Direct conflicts are detected when a developer checks-in their changes. Indirect conflicts are detected even later, during a build failure, an integration failure, or as a bug after deployment.; This dissertation presents our approach towards mitigating the effects of software conflicts by: (1) enabling developers to detect emerging conflicts (both direct and indirect), while conflicts are still small in size and relatively innocuous, and (2) thereby providing the developers the opportunity to self-coordinate to either avoid the conflict altogether or resolve it easily.; The contributions of this dissertation are threefold. First, it frames our approach in a novel framework that provides a historical perspective on the evolution of coordination technologies and the underlying coordination paradigms that these technologies follow. Second, it verifies the technical feasibility of our approach through the implementation of our research prototype, Palantir. Third, it presents quantitative evidence of the benefits of Palantir in the detection and the resolution of a larger number of conflicts through controlled user experiments. Specifically, our experiments show statistical evidence of the benefits of Palantir in facilitating users to: (1) detect conflicts early, (2) detect and resolve a larger number of conflicts, and (3) self-coordinate amongst each other.