Re: 'svn ls' slow over ra_dav

Ben Collins-Sussman <sussman@collab.net> writes:

> * A number of svn developers have "picked away" at the situation,
> trying to reduce the number of turnarounds. We do some
> client-side caching (ick) to try and help. Here and there, a
> developer has discovered a way to do one less PROPFIND. But, as
> you've noticed, we've not made a real dent in the problem.

As far as I can tell (and I'm no HTTP/DAV/DeltaV expert :) the "svn
ls" problem is nothing to do with the number of PROPFINDs, it's about
the time taken to process one particular PROPFIND. When I first
noticed the problem (back in issue 773) I had a look at the code and
it's complicated, PROPFIND is implemented partly in mod_dav and partly
in mod_dav_svn. I didn't put in enough time to understand the code
and I don't remember the details now, but one module calls the other
and the second module calls back into the first. Given the non-linear
behaviour I suspect something needs to be cached, or not processed
repeatedly, somewhere in mod_dav/mod_dav_svn. However I have no idea
what that "thing" is, or even whether fixing it is possible given the
current mod_dav implementation.

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org