I've started keeping all my files in perforce. So far it causes me only pain and helps me not at all.
I'm convinced now that game industry crunch is caused almost 100% by bad management. And my "bad" I don't mean "hasn't read Peopleware" bad, I mean completely irresponsible and juvenile bad. Basically what happens is this : up until 3 months before ship, management pays very little attention to whether the game is actually getting done; the only way they participate in the process is by diverting people onto tasks that aren't really important, or onto personal fetishes, or needlessly changing major facets of the project. Then, when you get within that 3 month end period they final go "oh shit" and get responsible and start making the people in the pit actually finish it. Now, I'm not saying the people in the pit are all responsible and good, they will also screw around and not wrap if left to their own devices - but the point is it's not their job to supervise themselves! If the people on the floor are not doing solid, directed work towards wrapping the project - it is the fault of the management.
It's tough being a lead sometimes, because I want to be all friendly with my team, but sometimes they just really fuck up and do stupid things or waste their time on unimportant tasks that they weren't supposed to be doing. Then I have to lay down the law, and noone likes that. It would be easier if it were just a situation of "do what I tell you", but it's not, I try to encourage some freedom and let people figure out their own issues, and it's tough when you have that mixed with straight task-mastering. People can take direction very comfortably when they don't feel invested in the work at all, but when they become personally attached to the work they become very resistant to direction.
Sadness. My Galaxy3 meshing stuff was probably the best in the world a few years ago (it may still be, even though I haven't really touched it). In 2001 I could render models that looked almostly exactly like the 1-million-poly original, but with 1-4k polys. There's no great invention there, it's all obvious stuff, VIPM + normal maps = good. But noone did it, still noone has done it, though we are getting very close finally with Doom 3 and all the copy-cats. My PPMZ and LZP and WaveCode coders were the best in the world at their specialties, but were never really used for anything. Then I stop maintaining them, and they get old, and they're no longer the best, and so now they're just totally useless historical curiosities. The code isn't even clean enough for them to be good learning platforms.