Resolving diverged branches in Subversion repository

[Jelmer Vernooij]

On Wed, Jun 18, 2008 at 02:56:11PM +0100, Russel Winder wrote:
> I can see this argument is going to run and run :-)

> On Wed, 2008-06-18 at 15:11 +0200, Jelmer Vernooij wrote:
> > Am Mittwoch, den 18.06.2008, 07:17 +0100 schrieb Russel Winder: 

> > > 1.  Subversion repository stores a Bazaar branch.
> > > 2.  Bazaar as intelligent Subversion client.
> > I would argue that these two points describe the same thing. 
> I'm afraid I cannot agree.  Certainly 2 can be covered by 1 (I am doing
> it now) but the problem is all the metadata that 1 requires that 2 does
> not.  The time taken to sort out the metadata really is quite
> significant.  And then there are the commit reports with 30K of data
> about stuff that is completely irrelevant for people in mode 2.
What I mean is that in git you can only import from and export to
Subversion. You can't use a Subversion repository directly from
git-svn and that's why I think the term client is misleading. 

> > IMHO git-svn doesn't work like a Subversion client, I would argue that
> > git with git-svn is a tool to interoperate with Subversion from git
> > (which is a different thing).
> I think we are agreeing and saying the same thing in different words.
> In the end though because git-svn does not store metadata it is a far
> better Subversion client that bzr-svn.  Sadly.
Yeah, we do indeed agree about the situation - just not the terminology.

I've started working on a dpush command to allow metadata-less push (instead
rebases your local branch on the fly after you push, like git-svn's
dcommit does). It's in the bzr-svn/dpush branch registered on lp.

Cheers,

Jelmer

--