[Python-Dev] Looking for VCS usage scenarios

[Stephen J. Turnbull]

Nick Coghlan writes:

 > If it isn't already there, suspending work on something to work on
 > something else would make a very nice scenario to cover, as it is
 > something even the core devs sometimes have to deal with.
 > 
 > I'd expect any DVCS to be able to handily beat what I currently do
 > with SVN:

Sure, but that's just you.<wink>  AIUI the "quilt" tool that kernel
people use simply automates exactly that process for you in svn or any
other VCS.  (I don't use quilts myself.)  The various DVCSes also have
specialized tools for managing patches.  Darcs is patch-oriented from
the get-go, ditto Arch.  Bazaar has "looms", Mercurial has "queues".
The advantage of looms and queues is that the patches are managed with
the VCS; I don't recall whether there's a way to do that with quilts,
but I presume so.

However, I suspect if you're expecting to do this more efficiently in
a DVCS, you'll be disappointed.  I'll make this scenario concrete in
the PEP (for git), but I'd say that the "operation count" would be
perhaps higher for me, using git as I like to, than in your scenario.
Of course they would be VCS operations, not using patch and diff.
What doing it in a DVCS provides is (1) a VCS record of what you did,
(2) which can easily be published, and (3) the option of using the VCS
to merge the two tasks later.

In other words, *D*VCS is about keeping and (optionally) communicating
*local* history, not about reducing the number of VC-related operations
you do.