CVS2SVN and bitching about converted repository size
I’ve seen a lot of people complain about the size of SVN repositories. The one thing all these people share in common is that they are using repositories converted to CVS2SVN. There is a reason for this:
CVS2SVN has a habit of misidentifying entire-branch copies. As a result, you will usually end up with branches that are as big as the original branch (on-disk size wise), rather than being , say, 1000 bytes.
This is not to say that SVN repositories are great at space usage. But as a rule of thumb, proper SVN repositories are always smaller than the equivalent CVS repository.
SVN has some problems in terms of repository storage formats. We know this. This is why SVN will probably going to move to something like revlogs for 2.0. But a lot of the specific size problems i’ve seen complained about on the web or mailing lists, are usually just bad conversions.
May 16th, 2007 at 1:59 pm
FYI: I fixed this problem last year and sent the cvs2svn developers a patch. It’s taking them a while, but they’re still in the process of integrating it.