This is mostly a bookmark/thought for later: see the current cvs2rss package and create something similar that works in a push fashion placing the data in a data structure with each commit rather than scanning the entire tree. (See examine_cvs OI package for an example of doing this…)