[darcs-users] revision identifiers

[David Roundy]

On Fri, Mar 28, 2008 at 06:56:20PM +0100, Declan Naughton wrote:
> On Fri, Mar 28, 2008 at 6:48 PM, Declan Naughton
> <piratepenguin at gmail.com> wrote:
> > On Fri, Mar 28, 2008 at 5:58 PM, David Roundy <droundy at darcs.net> wrote:
> >  > On Fri, Mar 28, 2008 at 05:26:27PM +0100, Declan Naughton wrote:
> >  >  > Why can't checksumming the context produce a suitable identifier, I
> >  >  > must ask? What are the uses in it being "universally unique" that we
> >  >  > would miss out on, anyhow? We can still seek through the repository
> >  >  > history and get specific revisions?
> >  >
> >  >  The checksum of a context wouldn't allow us to determine what context it
> >  >  describes, since the number of possible contexts we'd need to try goes as
> >  >  N!, where N is the number of patches in the repository.  I think that'd be
> >  >  a bit too expensive.  Best to just include the entire context.
> >
> >  We don't need to generate a sum for every possible context, just one
> >  sum for the actual context at every point in time (every time a record
> >  takes place).
> 
> Wait. What I mean is to associate the sum with the context, in the
> repo, and it will know the context that the sum identifies.
> 
> (and it would not be a good plan to save a context per record, but to
> simply mark the points in the history with the identifier, and the
> context can be generated from that. And if the history is
> restructured, then we can cater for that too without too much bother,
> I think)

What you're describing is synonymous with automatically tagging with each
record.  This is a reasonable option, as an earlier poster mentioned, but
it does have downsides (as also mentioned).  That is to say, this is what
your proposal would become, once you've catered to restructuring of the
history, and you've made it so that your checksums will work when used with
another user's repository.
-- 
David Roundy
Department of Physics
Oregon State University