On 1/30/07, Daniel Rall <dlr@collab.net> wrote:
>
> On Tue, 30 Jan 2007, Mark Phippard wrote:
> > Dan wrote:
> > >Would this make it possible to tie Eclipse's Change Set view into
> > >Subversion's changelist API (that you mention in "Message-ID:
> > ><ff892c5d0701170842i3454a39bye1f0f45f2b8a2e18@mail.gmail.com>"), such
> > >that the glue by which Subclipse "drive[s] the view" can also drive
> > >Subversion's changelist API?
> > >
> > >This way, Subclipse could maintain its own database of changelists (if
> > >necessary), but also interact in the same WC with other Subversion
> > >clients.
> >
> > Sort of hard to say. It would be very easy today to just relay the
> > add/delete file to changelist code that we have in place through the SVN
> > API. In other words, still use our own mechanism to track the
> changelist,
> > but also use SVN API so that WC contains same info. The problem would
> be if
> > we did not rely on SVN to also tell us the changelists and what is in
> them,
> > then we would not be aware of changes to the changelists done outside of
> > Eclipse. So realistically, I think we would want an all or nothing
> approach
> > and without trying it, it is hard to say whether it would work well or
> not.
>
> I see. So, ideally the changelists themselves would be sotred in the
> WC, and any additional information used by Subclipse would piggy-back
> off the changelist names in separate storage.
Yes, that would be the ideal. Store as much as possible in Subversion WC,
and use our own for anything else we needed. This would allow the user to
move between SVN clients without losing the basic information. I suspect if
the status API was modified as discussed, that could cover it.
--
Thanks
Mark Phippard
http://markphip.blogspot.com/
Received on Wed Jan 31 02:08:10 2007