On Tue, 2010-01-19 at 20:05 +0530, Kamesh Jayachandran wrote: > On 01/19/2010 07:23 PM, Mark Phippard wrote: > > On Tue, Jan 19, 2010 at 8:49 AM, C. Michael Pilato<cmpil...@collab.net> > > wrote: > > > >> Mark Phippard wrote: > >> > >>> Any idea how it works? CHECKOUT is serviced by the proxy server > >>> right? I would think it would lead to the same problem? Maybe the > >>> code winds up creating some kind of floating HEAD link as opposed to > >>> tying it to what it determines to be the HEAD revision. > >>> > >> CHECKOUT is serviced by the master, as it is part of the commit process. > >> CHECKOUT in WebDAV isn't a read operation like "svn checkout" is. It's > >> more > >> like "prepare to be modified" or somesuch. > >> > > Thanks. I knew this was a DAV checkout, but I thought it was also > > used in some read operations. Anyway, that probably explains why it > > works. > > > > Nice suggestion. > > > Committed in r900797.
Looks like a very good fix - doesn't make the proxy or the clients have any special knowledge. I have added issue #3561 <http://subversion.tigris.org/issues/show_bug.cgi?id=3561> to track this (mainly because I think it is good to have a place of reference for every significant bug fix that is made in a back-port branch, and also because it is not very easy to trace the emails in the Apache archive). - Julian