Ivan and I just spoke in person, and the summary of what I described is: using "HEAD" for the revision number opens us up to problems. If we started checking out !svn/rvr/HEAD when that was r10... but halfway through the checkout, we might start fetching r11 resources. We need a *consistent* revision during the entire checkout process.
If we were grabbing individual resources, then it would make some sense, but Ivan noted that we simply use ROOT/path/to/resource for those kinds of calls (ie. we avoid the !svn trees and go straight to the head resource denoted by the full URL). The motivation behind this idea for rvr/HEAD was because some actions fetched "baseline info" and used this form, and (seemingly) without any good benefit. Ivan is going to look into avoiding those fetches, rather than creating a new rev=HEAD concept. Cheers, -g On Fri, May 20, 2011 at 08:09, Ivan Zhakov <i...@visualsvn.com> wrote: > Hi, > > Currently in HTTPv2 enabled server we have various special stubs URL > that client can construct without contacting the server. One of them > is revision root: '!svn/rvr/REV/' [1] > REV can be only valid revision and in many cases client still has to > request HEAD revision to construct URL. > > So, is it makes sense to allow HEAD for revision number to be able > construct revision root URL without contacting the server? > > [1] > http://svn.apache.org/repos/asf/subversion/trunk/notes/http-and-webdav/http-protocol-v2.txt > > -- > Ivan Zhakov >
