RogerBW's Blog

I do actually like git. I find it needlessly obfuscatory and deliberately confusing in its syntax and terminology, but it basically does its job reasonably well. However, there are some popular blatant untruths that I think people would do well to know about.

"Git is distributed". Well, up to a point. You can poke back through the history (at least the history that hasn't been rewritten) in your local copy, without contacting anywhere else. That's lovely.

However, the use case I had for it was: I have a bunch of files that I work on from different places (e.g. desktop and laptop). When I start work on the laptop, I want to pull in any updates I've made on the desktop since last time I used the laptop, and push them back to the desktop when I've finished. Now in theory you can git pull in this way, but in practice you get all sorts of warnings about overwriting, and what you are supposed to do is… have a central git repository that isn't used directly, and all the others check in and out of (all right, push to and pull from).

"Git can track files through renames." Ha ha ha ha ha. Even with the git mv command, the file that appears after renaming is a completely fresh file. Git will probabilistically compare its name with the names of other files that disappeared in the same commit where it appeared, in the hope that one of them might be the same one. But it doesn't want to track the changes; git log simply starts from the rename unless you insert --follow, and I haven't been able to persuade git diff to produce anything other than arcane error messages.

I do actually like git. It's by far the best version control system I've used. Really. I just wish its users wouldn't lie about what it does. (Just like "org-mode can do dependencies" - no, not of the form "A depends on all of B, C and D, and C and D both depend on E" it can't.)