On Tue, Sep 04, 2001 at 02:51:32PM -0400, C. Scott Ananian wrote:
> On Tue, 4 Sep 2001, Alan Shutko wrote:
> > Branko �ibej <brane@xbc.nu> writes:
> >
> > > All such tools should use the svn command line directly. Unlike cvs,
> > > svn does (will?) provide all the information needed by such tool.
> >
> > Forking svn when opening every and any file just to see if something
> > is under revision control is not my idea of fun.
>
> I believe that's why the 'command-line inteface' mode of the svn client
> was proposed. It's like FTP; you open one session and you (or your
> program) types commands at it.
>
> Of course, when I looked at implementing this, the client library of SVN
> was architected completely wrong for this; there wasn't a way to
> 'authenticate once, use multiple'.
IMO, the concept is bogus, and I'm not sure where Branko came up with the
"planning to do this". If something wants to do a long series of commands,
without forking each time, then it can simply bind to the library and do it
that way. That could be a Perl, Python, or whatever script (based on the
SWIG bindings to the client lib).
Regarding Scott's auth once issue: that is probably unavoidable for the
*client* library. If you wrote a script directly against RA, then you could
do a single auth. But that would also be a lot more work :-) The issue is
that each client lib function is going to open a new RA session.
Maybe it would be possible to revamp the client lib API to include an RA
session, but that is probably a post-1.0 issue. (altho patches are welcome:
if it isn't too disruptive, then no reason it couldn't go into 1.0 if
somebody applied their time)
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:40 2006