On Mon, 2005-11-21 at 17:02 -0500, C. Michael Pilato wrote:
> Daniel Berlin <dberlin@dberlin.org> writes:
>
> > On Mon, 2005-11-21 at 11:48 -0500, Marc Sherman wrote:
> > > Daniel Berlin wrote:
> > > >
> > > > Honestly, I don't care what we do in mod_dav_svn. I only bothered to
> > > > add svndiff1 support there for completeness. I have zero interest in
> > > > DAV, as it is complicated and slow :)
> > >
> > > Just out of curiosity, is that a personal opinion, or is it widely held
> > > by the dev team?
> > >
> > Personal opinion, but one held from watching DAV in subversion :)
> >
> > The only way to make it fast is to use custom reports.
> > But if the only thing that knows how to interpret your custom reports is
> > Subversion, being custom and all, then uh, you haven't won anything by
> > using DAV :)
>
> The benefits of DAV (such as autoversioning support, browsable
> repository URLs via directory GETs, etc.) are not lost by allowing
> mod_dav_svn to communicate solely via customer REPORTs when it's a
> Subversion client doing the talking.
>
> We could start using custom REPORTs and POSTs for all Subversion
> operations, but additionally support the "real" DAV stuff necessary
> for those aspects of DAV we care to shoot for. In other words, start
> with a working, performant version control system, and then add
> cheap(ish) goodies as desired.
>
What advantage does this have over tunneling something sane over HTTP,
and then just having a small dav implementation that supports the other
stuff?
Most of the hard to maintain code is in the custom reports and whatnot.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Nov 21 23:09:49 2005