On 5/31/06, Peter Scmsvn <scmsvn@gmail.com> wrote:
> Hi,
>
> regarding that thread, It seems that tunnelling svn:// loses all the author
> information.
> (which is logical, since anonymous login has to be allowed.)
>
> So I'm still stuck with the problem that using svn+ssh:// does one handshake
> for each
> transaction made, including opening folders, which takes approximately 2
> seconds, that can
> be a little annoying...
>
> Another solution would be to use apache+ldap but woud it be possible to
> chache the
> connection information and keep alive the ssh session, as long as, lets say,
> the repo-browser is opened ?
>
> Thanks,
>
> --Peter.
>
>
Pretty sure you're asking this in the wrong place. This is something
the subversion libraries would need to support. Also apache+ldap
would suffers from the same problem. It re-opens and re-auths multiple
times just like svn+ssh. However, your comment on svnserve is wrong.
It does not require anonymous access.
http://svnbook.red-bean.com/nightly/en/svn.serverconfig.svnserve.html#svn.serverconfig.svnserve.auth
mentions how to disable anon-access.
I'd love it though if they could redesign things to not have to
continually re-auth, as this was one of the things i had to overcome.
Our initial server setup was apache+mod_pam+winbind, but winbind was
using way too much CPU time to authenticate and was making operations
like checkout, or log retrieval, take 6x longer than no
authentication. Ended up using apache+ldap which used less CPU, but
still because it has to authenticate every time, it takes about 30%
longer because of the auth. Didn't use svnserve because tying it to a
windows domain would have been much more complicated, and at it didn't
support per directory permissions.
- Jody
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tortoisesvn.tigris.org
For additional commands, e-mail: users-help@tortoisesvn.tigris.org
Received on Wed May 31 17:01:57 2006