Distributing Distributed Revision Control Systems

Abstract

Current revision control systems are commonly used in a distributed fashion but rely on centralized stores and low-latency communication. In this work we evaluate how they would fare in fully distributed high latency networks such as delay tolerant networks (DTNs). We show that current revision control systems impose significant costs under these conditions even in moderately-sized networks. By simplifying/improving the merging process, these costs can be reduced. We also show that speeding up or slowing down communication can reduce the costs significantly.