30 Years ago...

> The main reason Git took off outside the Linux kernel community is that ordinary programmers thought that if they just used Linus Torvalds's uber-hacker tool, they would themselves become uber-hackers like Linus Torvalds

I doubt this was a serious reason. Git performed much better for a while (I haven't compared them recently) and large influential (at that time) software projects including Ruby on Rails and later Android etc switched to Git. GitHub took off making adoption much more easier.

From what I remember, git was much less opinionated at the time. We had some hg repositories internally and I ended up using git-hg to interact with them because I found mercurial so frustrating. If you were working on a few small patches at the same time you needed something like the mq extension which gives you sort of 2nd-class commits, with a completely different API. As soon as you pushed something to a server the commit got marked immutable, so you could never setup a buildbot to try-build your patches.

I never understood the point of "immutable changesets". It feels like working with one hand tied behind your back.

Git is a "stupid content tracker" and thus made it possible to fit it to a workflow that worked for us, rather than what the tool wanted. For me git vs hg was like night and day. Also, the git index is IMO was a real innovation, that you could build commits piecewise.

Looks like hg is less strict these days, but once you've got everything to git the barrier to switch back is huge.

FWIW I started using git, and not the other alternatives, because I'd heard of git and not the other alternatives :) After all, Linus Torvalds has quite a big profile.

"The main reason Git took off outside the Linux kernel community is that ordinary programmers thought that if they just used Linus Torvalds's uber-hacker tool, they would themselves become uber-hackers like Linus Torvalds."

No, that was not the case.

> The main reason Git took off outside the Linux kernel community is that ordinary programmers thought that if they just used Linus Torvalds's uber-hacker tool, they would themselves become uber-hackers like Linus Torvalds.

My recollection of the events (I was in the bzr camp) was that people switched from hg/monotone/arch/bzr to Git because

1) Git was blazing faster (milliseconds vs seconds for `$VCS status` = it can be integrated in PS1)
2) Git was less opinionated (just use whichever branch workflow you want)
3) Git was more tolerant of errors (git rebase).

Point 3 was especially important for me. Git rebase gave you the ability to polish your final results, instead of forcing you to show "how the sausage was made".

Monotone used file-level SHA-1 IDs, so it had way more overhead than git. Each file required a separate operation to sync. It also used SHA-1 for the whole repo for a version, just like git.

There were some obvious optimizations that were missed, but the development of Monotone more-or-less died after git.

It also turned out that an SQL database for VCS data storage was not a good idea in general, because of all the overhead it brings compared to plain files.