Daniel Carrera writes:

 > I was hoping that we would some example to justify the claim that the 
 > theory of patches is somehow better than 3-way merge.

Well, that's what I've been asking for, too.  If I had seen a
concrete, easy to present example, believe me, I'd dig it up.

I think one hand-wavy thing is that Darcs has access to all of the
history between the common ancestor and the divergent branches.  So
Darcs can *compute* the offsets and fuzz (well, Darcs hunks can't
fuzz, but something like that maybe?)  That clearly beats patch.
Whether the additional information allows us to beat 3-way merge, I'm
don't know.

I think one important psychological aspect that hasn't been mentioned
yet is that sometimes you think ahistorically, in terms of patches
(changesets, features), and sometimes you think historically (in terms
of revisions and branches).  And most people are biased one way or the
other.  I'm most active as a release manager; history is very
important to me, while I hate patches and features because I've never
seen a feature that didn't bring more than its share of bugs. ;-)  So
I think historically-oriented people, even folks who are maintaining
stacks of patches a la Andrew Morton (who invented 'quilt'), are
likely to favor a snapshot-oriented system that allows explicit
manipulation of the history DAG.  Ie, git.  While people who think in
terms of "recombinant features" and don't manage releases are likely
to prefer changeset oriented systems like Darcs and (original) Arch.

If managing patches are the way you think about version, then git is
going to make no sense at all, and Darcs is peaches and cream.  If
traversing a DAG is your view of how the world works, then git is the
obvious way to go, and Darcs's focus on patches seems quaint at best.

I don't think you dare phrase it this way in front of those who
haven't had a couple doses of the electric Kool-Aid, but you could say
that Darcs even provides an "algebra of features".

_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users

Reply via email to