[darcs-users] patch file naming

[Ganesh Sittampalam]

On Fri, 19 Mar 2004, Kevin Smith wrote:

> I am not an expert. Based on a few minutes of thinking, my current
> suspicion is that we can think of a patch as having two aspects: WHAT
> and WHERE. I am guessing that the WHAT cannot change as a result of two
> patches being commuted, but the WHERE will often change.

If that's true, then perhaps one could work out some kind of "minimal"
source tree that such a patch could possibly apply to; a canonical
representation of the patch would then be the minimal source tree + the
result of applying the patch to that tree. Of course, it's not at all
clear what filenames to pick for such a tree.

Indeed, in general patches that do two different things on a given tree
could actually be turned into each other by commutation with other
patches, so a completely valid canonical representation is impossible. It
may still be possible to get something that's "close enough", though.

Cheers,

Ganesh