About git-thought [was: brisbane: initial cut at a mergeline cache]

[Andrew Bennetts]

Stephen J. Turnbull wrote:
> Andrew Bennetts writes:
> 
>  > Why would you be numbering revisions that aren't in your ancestry?  To
>  > describe a pending merge?
> 
> There's that.  But more often for me, to describe branch divergence.
> Obviously, there's a possible merge in the future or I wouldn't
> *care*, but it's not pending in the sense of "as soon as I can grab
> the lock".
> 
> This is why your conversations with git aficionados don't go anywhere
> fast, I think.  You never saw a branch that didn't look like a pending
> merge.<wink>  But we see branches as a cache of searches through
> revision space.  Sometimes we're looking for a way to converge.  But
> often we're looking at ... something else.

Oh, I certainly understand the desire to have a coherent view of N divergent
branches as one unified graph.  It's something I want to do from time to
time after all, and AIUI qlog and viz are moving in this direction if they
aren't there already.

Sometimes I want to see what a merge would look like as a diff, sometimes I
just want to browse the part of the revision graph unique to one branch,
etc.  (By “browse” I mean roughly “see commit logs and general activity”.) 

Or more broadly perhaps, to understand the contents of one branch relative
to another, or even relative to a set of other branches.  This is the sort
of thing I'd love to have good tools for, probably a GUI.

> If now you ask "what?", then you still don't get it.  There are lots

I don't know.  You're being awfully abstract... I understand you're talking
about support for a class of use cases rather than any particular use case,
but it doesn't make it at all clear if the class I have in mind is the class
that you have in mind.

[...]
> Why should you care?  Well, some of us would like to be able to do in
> bzr, too, what is simple in git.  As long as it's not even thinkable,
> though, we can't think about it, and we won't be considering bzr as a
> viable alternative to git.

Yes, we're agreed about this.  If there's something that bzr should make
easy that isn't easy yet, we'd like to fix that.  Concrete suggestions are
particularly helpful for making this happen <wink>.

So, how would you like data-not-in-a-particular-ancestry to be labelled (and
what aliases would you like to be able to use to refer to them)?

-Andrew.