Changes at Subversion Revision Control System
Jon Trowbridge and Tony Zale have blogged about changes recently implemented at the open source revision control system, Subversion.
The Subversion team have added a custom replication system based on the Paxos algorithm for reaching a consensus in a network of unreliable processors. Now, whenever the user makes a change to a repository, the new data is copied to several different data centres before Subversion reports that the commit has succeeded.
The team have also rebuilt Subversion’s HTTP-based protocol with an emphasis on optimising this service for browser access. New caching headers have been added, which can be manually turned on and off using the google-cache-control property.