You say that "Subversion was not that great of an idea" but if you go back to 2001 (which is when svn became selfhosting), or even 2004, CVS was such a disaster for some use cases that the need for a replacement was readily evident. Add to that the fact that svn was a compelling replacement for Visual Source Safe, and it was definitely a worthwhile idea. Sure, it's now the "old safe standard" of version control, but there's a lot of value in that for many companies. Several of the original svn team (including these two) have admitted that some of the design decisions (the handling of tags and branches most notably) were probably mistakes, given the value of hindsight. They got a lot right too, though. The delta editor interface is nicely abstracted to the point that hgsubversion uses the same delta editor as the svn command line client to do something the original design spec never would have anticipated. Yet it works extremely well.
Subversion is much simpler for people to "get" than other systems. I've been a heavy evangelist of DVCS tools and workflows for a while at work, and it's taken since October (8 months) to get any critical mass behind actually even investigating switching to those from SVN.
Let's face it: Subversion is many many times better than nothing, which is what a ton of people do. Most college graduates in CS or CSE come out of school never having touched version control in my experience, and the ones that have seen it were exposed via open source, not their coursework. Evangelizing Subversion is (really lousy analogy warning!) like convincing people to smoke cigarettes instead of doing illegal drugs. At least it's less bad for them, and the barrier to entry might be a bit lower. When we had a poll of the local Python users group one meeting, "email" was the most popular version control solution, followed by a basic dead heat between svn/bzr/git/hg. The fact that people still view email as a valid answer is far more troubling to me than evangelizing an old and very mature tool instead of one of the new upstarts.
I think SVN is a classic case of overreacting. "CVS is bad, and it has first class branches and tags. Those must be bad too!" The result was a very poor design ("everything is a file!") and a very poor implementation. The implementation can be fixed (and it is going in that direction), but the design is broken to the point of being unusable.
I am glad that CVS died, but it was "replaced" with something not-much-better.
Subversion is much simpler for people to "get" than other systems. I've been a heavy evangelist of DVCS tools and workflows for a while at work, and it's taken since October (8 months) to get any critical mass behind actually even investigating switching to those from SVN.
Let's face it: Subversion is many many times better than nothing, which is what a ton of people do. Most college graduates in CS or CSE come out of school never having touched version control in my experience, and the ones that have seen it were exposed via open source, not their coursework. Evangelizing Subversion is (really lousy analogy warning!) like convincing people to smoke cigarettes instead of doing illegal drugs. At least it's less bad for them, and the barrier to entry might be a bit lower. When we had a poll of the local Python users group one meeting, "email" was the most popular version control solution, followed by a basic dead heat between svn/bzr/git/hg. The fact that people still view email as a valid answer is far more troubling to me than evangelizing an old and very mature tool instead of one of the new upstarts.