[Kde-scm-interest] Have we arrived to a dead end?

[John Tapsell]

On 17 February 2010 23:23, Oswald Buddenhagen <ossi at kde.org> wrote:
> On Wed, Feb 17, 2010 at 04:50:52PM -0600, Boyd Stephen Smith Jr. wrote:
>> Enhancing git to support narrow clones and keeping everything in one
>> repository lets individual contributors decide which part(s) of the
>> tree are important to them, while allowing atomic cross-module commits
>> (since they are all in the same repository) and trivial
>> copy/move-with-history (no special subtree merge required).
>>
> that's nice in theory ... in practice, a repo with the size of KDE/
> would be just plain incredibly slow to work with. in git, status and
> commit operate globally. imagine the equivalent of "svn st KDE/" each
> time you want to commit something. less obviously, this also has a
> tremendous effect on the speed of rebasing and other more advanced
> operations. just compare qtcreator and qt - just to name one "fairly
> big" and one "pretty huge" git repo.

Agreed.

I think having kdelibs in one git repository, kdebase in another etc,
is acceptable.  It doesn't need to be split up any more than that.
I use git-svn with the whole of kdebase.  git status takes a while the
first time you run it, from cold.  And I have a very slow laptop.

Specifically:

$ time git status
real    0m14.884s

$ time git status
real    0m0.269s

So after the very first time you do it, for the rest of the time
everything else is pretty much immediately.

John