Stefan Sperling wrote on Thu, Nov 28, 2013 at 13:03:59 +0100: > On Thu, Nov 28, 2013 at 01:52:08PM +0200, Daniel Shahaf wrote: > > Philip Martin wrote on Wed, Nov 27, 2013 at 22:21:00 +0000: > > > Not sure I follow that. The client changes svn:date before the > > > pre-commit hook but the pre-commit cannot currently distingush the > > > initial svn:date set at txn create from an svn:date set by the client. > > > We need some indication that the client has explicitly set svn:date, > > > including deleting it to indicate a revision with no date. > > > > > > > That indication could be the client's authenticated username being > > "svnsync". > > I don't like this idea. It doesn't seem right to enforce a specific > username for protocol purposes.
The book recommends putting an author check in start-commit. > Forcing people to use a specific username is awkward with ra_local and > svn+ssh. I'm not sure what the problem you see is. Even with svn+ssh you'd get here the username that the sync job authenticates as, not the OS username. And if people run svnsync on ra_local, they could just have their hook set the revprop unconditionally regardless of $AUTHOR.
