On Mon, Jul 29, 2002 at 11:03:08AM -0500, Karl Fogel wrote:
> We're interested, yes. To some extent, this duplicates work we've
> already been doing on a "cvs2svn" converter, but that's okay.
Hopefully, you'll also pick up a p42svn converter, a vss2svn converter,
etc :). Not quite for free, but hey...
> For big repositories, direct-to-backend is going to be a real win.
> We're doing it that way too.
So far, I've leaned towards the "hey, I just want to install this on
<name-your-OS-here> and run with it" user, so wrapping each SCM's
command-line tool is the current approach. Backend access will often
require some sort of C compiler and the guts to use it (cvs is a rare
exception), but offers speed, accuracy, and (perhaps) correctness
improvements.
> > FWIW, VCP so far has been funded by Perforce thus far and Perforce is
> > quite interested in seeing VCP and RevML used as a standard tool for
> > interchange between repositories.
>
> Wow. I'm tempted to ask how you handle discrepencies between
> repository schemas, but then again, you did point us to a theory link.
Not that kind of theory, though. So far, the schemas map pretty well.
There are a few funny areas that don't seem to matter much, like the
different semantics for labels/tags, although don't get me started about
VSS. Luckily, I expect most people doing VSS->whatever conversions to
have seen the light and expect some duct tape to be used to work around
VSS' braindeadness.
I'm working out the mapping of branches now.
- Barrie
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Jul 29 21:58:52 2002