[darcs-users] Re: [darcs #368] Proposal: unique repository ID

[David Roundy]

On Mon, Jul 04, 2005 at 11:36:39AM +0100, Jamie Webb wrote:
> On Mon, Jul 04, 2005 at 11:28:48AM +0200, Stephane Bortzmeyer wrote:
> > As you wrote, collision would just be a nuisance, not a showstopper so
> > I believe my idea still hold
> 
> Mmm, but your idea will cause collisions to be guaranteed for every
> subsequent patch in a pair of repos. That makes the scheme completely
> useless from then on.

I think this is a pretty strong argument against the idea of unique numeric
repository IDs.

> [...] OTOH, Collisions in patch hashes just mean that you can't refer to
> those particular patches by hash. The rest of the repo can remain
> unaffected.

Yeah, I think that hashes of the patch ID can serve as sufficiently useful
unique patch IDs.  The patch ID (author+date+name+log) is already required
to be unique--although darcs doesn't do anything special to enforce this,
darcs fails if it's not true, so we've already got a unique patch
identifier.  The patch hash is also already used, and in the current
implementation also required to be unique (since it's the patch filename).
It's also used in parts of the user interface (via --match), it's not not
easy to obtain.  I think the outstanding issue with unique and convenient
patch IDs is really one of making them convenient rather than that of
making them unique.
-- 
David Roundy
http://www.darcs.net