OSNews reports Subversion 1.4.0 Released. “This is a feature release of Subversion [Updated link], featuring BDB 4.4 and repository auto-recovery support, a new tool for synchronizing repositories (svnsync), major speed enhancements in the versioned filesystem and the working copy, and of course the usual host of bugfixes and minor enhancements. Additionally, check this article on how to Set up Subversion and websvn on Debian.”
Good timing! I've been using subversion for the past year on a web development project with another (remote) developer, and have enjoyed the power and flexibility of the tool, as well as some of the cool add-ons, clients and scriptability.
Now, it's time to consider moving existing projects out of Visual SourceSafe and into subversion. The folks at Pumacode offer an vss2svn tool that runs as a native Windows executable, written in Perl and C, with the source available under an open license. Pumacode tried an interesting tactic to convert the VSS repositories: rather than interogate the VSS binary to retrieve files, it reads the repository files directly and interprets the results from there. There are some advantages where older versions might be corrupted, or to retrieve files flagged as deleted, which they say VSS will not allow.
On a 2 Ghz Pentium-M with a gig of RAM, it took about 2 hours to process my current VSS repository, which consists of forty thousand files and around 1.4 Gb of disk space. (The authors of vss2svn caution that it's better to convert the entire repository than to risk further corruption by pruning it first; leave that task to subversion post conversion.) This generated a dump file of 850+ Mb. Transferring that to the Linux box with a new repository took a few minutes, and loading the data about 20 minutes. Using RapidSVN from the Windows box, I was able to browse the subversion repository and confirm that files and folders and log history comments look about right. I'll confirm by checking out projects of interest and diff'ing them against the current development copies.
I had anticipated a different tack, using COM Automation to drive VSS, as I described in Essential SourceSafe. As a learning project, I had proposed using Python to browse the repository via COM Automation and use the excellent Python-svn bindings to migrate portions of a VSS repository to subversion. I still plan to try that, and to compare-and-contrast the results between the two techniques, while I learn a little more Python.