Re: Debian Perl Group meeting at DebCamp - 2008-08-06
Gunnar Wolf <gwolf@gwolf.org> writes:
> Russ Allbery dijo [Fri, Sep 12, 2008 at 04:38:32PM -0700]:
>> Every place that I use Git, I use a separate repository per package. I'm
>> skeptical it makes much sense in the Git paradigm to have a single giant
>> repository for every module that the Perl group would package, for
>> instance.
>> Subversion tends much more towards centralization. Git repositories
>> are cheap, and there's usually no reason not to have lots of them
>> (although you have to write some quick scripts if you want to do mass
>> updates, and mass updates are a lot slower).
>
> I find it an advantage that our hundreds of modules are at one single
> location - And that they are _not_ on my hard disk ;-) Working with a
> package is usually quite cheap for me (i.e. I almost always work in
> /tmp, with very short-lived directories). Quite some time ago, I _did_
> have the whole tree in my home directory (including branches and tags),
> but... it was just too impractical in the long run. So... If for nothing
> else, this is a good case for subversion IMO.
I'm not sure what in there is a good case for Subversion, but I'm maybe
missing something?
With Git, you'd clone the repository of whatever package you were working
on, do some work, push your changes back, and then can just delete the
local repository. It's fairly similar to what you described for
Subversion. Assuming, that is, that each package gets its own repository,
but I think that's the most natural way to structure things with Git.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: