DeltaV conformance vs backward compatibility in fixing the issue 2151

Hi.

The previous patch for issue 2151(r18513) improves performance only
in 'svn ls -v' over ra_dav between both patched svn client/server.
There is another more significant fix for this issue.
See fix 1 in the following URL.
http://subversion.tigris.org/issues/show_bug.cgi?id=2151#desc11

The fix is filtering out DeltaV properties from allprop request.
This is also a way to get conformance with DeltaV. But this breaks
backward compatibility because old(and existing) ra_dav code is
expecting that an allprop response contains DAV:version-name and
DAV:creator-displayname as the last committed revision and author
respectively. I'm not sure but suspecting ra_serf has same expectation.

I think that filtering out DeltaV properties except above two is
a reasonable compromise. That improves more svn operation including
ls(with old client) and checkout. But large testing is still required
to investigate all similar wrong allprop usage.

I have implemented this compromise, but have not tested yet. I have
now started learning about the Subversion Testing Framework.
To continue working, I have questions:
  - Is this compromise reasonable and acceptable?
  (The following questions is valid only if the first one is true.)
  - Can I send a patch that surely requires large testing before making
    tested by myself?
  - Is there a chance to applying this to 1.3.x?

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