On Apr 30, 2005, at 9:22 AM, SteveKing wrote:
>
> Sure, I could do that. Most of the features I'd like to implement in
> TSVN could be done by calling several Subversion API's. The problem
> here is as always speed. svn_client_ls() is slow enough as it is
> (depending on server, but always _much_ slower than browsing it with a
> webbrowser). Adding another call which contacts the repository would
> make it so slow users would start complain a lot (they already do now,
> but there's nothing I can do about that).
>
Does that mean that you won't show remote locks in TSVN 1.2's
repository browser? I would at least make it an option!
Your point about network turnarounds is taken; my first instinct for
implementing svn_client_ls3() would be to call svn_ra_get_locks()
internally and add the information to the dirent2_t's it returns. But
then this would be just as slow as if you did this task yourself.
So the only way to efficiently get lock information is to really change
both network protocols, so that the lock info came back in the single
svn_ra_get_dir() network request. (Actually, the ra_svn protocol would
have to be extended here, but not DAV. ra_dav already asks for a list
of properties when it does a recursive PROPFIND; it just needs to ask
for an extra lock property.)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Apr 30 16:33:24 2005