Switch sql to git

[adehnert]

We're nominally now using git for sql.mit.edu. We should finish that switch:

• document (I guess in the locker) what the correct repo location is
• convert all shared checkouts to git:
  • servers (k-s and s-b)
  • locker
• make sure we can commit changes from the servers (password auth + smart HTTP? daemon.sql + AFS?)
• Sync any commits made to ​svn://presbrey or ​svn://sql.scripts to the git repository

[achernya]

It appears I've accidentally done about 90% of this by working on #342. I don't have a good story for managing the configuration of the two servers yet. I'm currently considering an XVM-style mako template in a config package for my.cnf...

[adehnert]

make sure we can commit changes from the servers (password auth + smart HTTP? daemon.sql + AFS?)

If we're down with Github, the "correct" answer may involve deploy keys (see ​https://help.github.com/articles/managing-deploy-keys, which appears to talk about deploy keys and some alternatives).

[adehnert]

Mitch would presumably like us to handle the requirements in Scripts #97, comment 3 as well, although given the current state of sql's repos, failing to do so isn't exactly a regression, I think...

[adehnert]

There's some discussion of our pony version control setup on -c scripts -i trac-#325 today (2013-06-27). I think we want to be able to push from the server (possibly only to a staging branch), and want to have reasonable author/committer somehow (a pre-commit hook to require a non-root value, or a bashrc that sets it from the GSSAPI principal, or something).