Jim Blandy <jimb@savonarola.red-bean.com> writes:
> I'm not especially fond of the way CVS does merges into the working
> directory, either. As you point out, it's not only that the
> repository doesn't store everything you wish it did --- it's that the
> working directory management is messy, too.
>
> You're not the only one who has expressed a wish for a more coherent
> way to manage merge participants than renaming the original to
> `.#foo'. But unless Karl has surprises up his sleeve, I think we're
> going to imitate CVS up to 1.0. Hopefully, we'll then have a decent
> code base to experiment with other merge disciplines.
The main issue is that all the ancestry involved in a merge is not
recorded by in CVS.
However, recording complete ancestry, and taking it into account
during merges and updates, is a well-understood problem (at least, we
thought we understood it some months ago), and the solution Subversion
plans to use is documented in the spec. In fact, it's one of the few
things in that spec that's not been obsoleted by code already. :-)
1.0 probably won't implement it, only because 1.0's goal is to be a
CVS replacement -- it'll get you out of Egypt, but it won't get you to
the promised land just yet.
Received on Sat Oct 21 14:36:07 2006