[Daniel Shahaf] > when the slave hasn't synced up yet, I assume (not knowing the proxy > code) that the editor would be opened against a revision N1, such > that on master branches/ has been touched in some revision N2 > N1. > Therefore, as far as master is concerned, the error is proper. > (consistent with the FS design)
Yes, but why is this only a problem with a write-through proxy? Shouldn't the same race exist between two clients talking to the _same_ master server? (Not just cp and mkdir, but anything that on the client side would have been considered a tree conflict.) I don't know the exact sequence used here - is there a server-side lock during the commit that prevents these conflicts? If so, the write-through proxy should really perform the same actions on the server that cause this lock to perform the same function. -- Peter Samuelson | org-tld!p12n!peter | http://p12n.org/
