Re: patchsets

       But you still have to remember to do it. I think a system genuinely
       tailored for patchset maintenance would prevent the user from even
       having to think about rip-driver-and-VM unless there was a conflict
       between them.

Putting a not-quite-literal filter on that, I more or less agree.
A workflow based UI for revision control is highly desirable.

The technology I described to you is ammenable to such a UI.

Chasing the CVS-ui-replacement with a not-quite-thought-out plan "is
not, specifically, good".

        That's the problem I was trying to identify, and I don't think
        arch solves it.

With the namespace and the merge-ops, arch gives you a solid
foundation for even thinking about the workflow based UI you are
vaguely imagining. More importantly, it demonstrates that such a UI
should not depend on a particular rev ctl storage manager (by
presenting the relevent abstractions in a storage-manager independent
way).

        (A layer on top of arch, or on top of Subversion or CVS, might
        solve the problem, although my experience with that sort of
        layer-on-top is that it works fine until something goes wrong,
        and then it fails in a confusing or catastrophic manner).

Oh come on now -- layered software is bad?

-t

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Oct 15 07:43:55 2002