Re: Top 5 things that aren't in Debian but should be :-)
On Tue, Jan 13, 2004 at 09:33:26AM -0600, Steve Langasek wrote:
> On Tue, Jan 13, 2004 at 12:28:49PM +0000, Colin Watson wrote:
> > *If* all its dependencies are ready in testing by its release date and
> > the maintainer acts reasonably quickly to get subversion 1.0 packages
> > into unstable, it seems plausible that it could be a late addition to
> > sarge, since it's new and high-profile. Of course, I'm biased, since I
> > don't want to try to maintain subversion-sarge packages in the same way
> > I've been maintaining subversion-woody packages. :-)
> > Any opinions from -release?
> Well, my biggest concern here is that subversion has broken its wire
> protocol several times over the past year. Are there indications that
> this state of affairs has improved? It doesn't make much sense to
> bother with subversion in a stable release, if our client is going to
> stop working with all the servers out there a third of the way through
> sarge's lifecycle.
Yes, this has improved. Pre-1.0, it was always documented that there was
no on-the-wire compatibility guarantee (well, you were guaranteed one
version in either direction for ra_dav), and during development
compatibility was sometimes broken deliberately. However, the whole
*point* of Subversion 1.0 is that protocol and API compatibility will
not be broken in the 1.0 cycle. In other words, it's to behave like a
shared library: any protocol incompatibilities are reserved for 2.0.
IME the Subversion developers are highly aware of compatibility: the
protocol breaks to date weren't "we're too incompetent to make this
compatible", but "we know this breaks compatibility but until 1.0 that's
OK, and we want to make sure the interface we finally release is the
best possible". I'd be truly amazed if we ran into any compatibility
problems with 1.0 final.
Colin Watson [firstname.lastname@example.org]