[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnu-arch-users] Re: darcs vs tla

From: Thomas Lord
Subject: Re: [Gnu-arch-users] Re: darcs vs tla
Date: Mon, 8 Nov 2004 15:27:31 -0800 (PST)

On the topic of darcs, especially the part of this thread talking
about the notion of commuting patches by variance adjusting and
building merge intellegence out of that ......

How much experience is there with the merging practices that result
from that approach?   On first principles I'm highly skeptical of it
because variance adjustment, unlike first-order diff and patch, has a
terrible relationship to program /semantics/.   In other words, while
diff and patch are strictly text based, nevertheless, on a bunch of
less formal, and admittedly less guaranteed levels, they do very well
for capturing changes to a C program in a form that (for practical
purposes) captures the /meaning/ of the changes.   So a diff on some
file is a not bad approximation of a logical statement of how the code
in that file changed.   But once you kick in variance adjustment, the
basis of the darcs commutative-patches hacks, i have severe doubts
about how well you're sticking to that lucky pun of textual diffs and
logical summaries of changes.

So: does Darcs' merge operators (the deep distinction it has from
arch) really "add value"?  at scale?   What's up with those?   Does
anyone actually /know/?

Next: suppose those operators /are/ just the bee's-knees and we sorely
miss them in arch.   Don't they map onto arch archives in some
interesting way?   I rather suspect that they do....


reply via email to

[Prev in Thread] Current Thread [Next in Thread]