Arch
Arch has gotten really rather impressive since I last looked at it. It’s design has stabilised enough to be rewritten in C (from sh), and the repository format seems remarkably sane and powerful, if perhaps a little overly verbose. Distributed repositories over plain http, ftp and NFS all seem supported, which is way cool. And hey, it’s already into its first rewrite, so it’s likely to not be accumulating loads of misfeatures. Subversion, meanwhile, is still answering the question “Is Subversion stable enough for me to use for my own projects?” with a less than reassuring “We think so!”.
Two theories leap to mind. One is that the activity in the version control arena is directly attributable to the obnoxiousness of Larry McVoy’s occassional rants on linux-kernel about people’s ingratitude for his very good, but proprietary bitkeeper. This is a corrollary to the “free software is about scratching an itch” theory: the more it itches, the more development there’ll be. The other is that subversion, by being more popular and buzzword compliant, attracts more of the fashion conscious, leaving arch with more competent people. A prediction: neither of these theories will stand up to analysis. But wouldn’t it be funny if they were true?
UPDATE 2003/09/01:
Martin’s had a look at arch too. His take’s similar to mine, and his major criticism seems to be that the docs for tla don’t match the behaviour of tla (AIUI they match the behaviour of larch — the shell implementation of arch, tla’s the C version — instead).