In the same way that you’d replace a shovel with a backhoe or carpool to work instead of riding your bike…. Taking an object-oriented approach of software configurations just makes sense. Inheritance on update!?!? . It’s brilliant! share changes implicitly across your peers/teams/globally, instead of manually. This entirely eliminates the mess of private branches and allows individuals to collaboratively dev/test in more realtime while still separating “ours” from “theirs.”

I’ve started consulting the stream-based approach to CM instead of hard branches. You could argue that clearcase config specs allow you to emulate streams… but it requires manual management of adjusting rules – hence the admin requirement. Meh. They are still on the old branch model. I think accurev’s model is the cleanest so far. if I am doing TDD but working along side other peers or teams, my daily ‘update’ will automatically incorporate their file changes… not just my local ‘feature’ or ‘branch’. so my TDD is more realtime in a global scope but in local context. In a large[r] environment, I can’t imagine working in any other model. $ rm -rf /usr/local/[branch-based-scm-binary]

What I want from an sccs is a) let’s say what an eclipse/cvs will offer with b) occasional, as in once quarter, freedom to do bizarre stuff one has to do thoughtfully and by hand.

Applications or house rules that force one to permanently live in the “b)” world are an abomination, and should be promptly dismissed. We can find a good job for that guy who administers it, in some other capacity. Hell, he’s proven he’s smart enough.