Re: [Monotone-devel] Re: Proposal for human readable revision IDs

[Nathaniel Smith]

Well, I've said it before, but... :-)

On Fri, Sep 09, 2005 at 08:08:35AM -0400, Hendrik Boom wrote:
> It was its totally distributed nature that convinced me. not the hash codes.
> I would have been happier with some locally unique version name that is
> combined with a globally unique tag identifying the site at which the
> locally unique name was created.  The reason is that I am still
> paranoid about losing emtire files by hash collisions.

"Globally unique tags identifying sites" are far, far more prone to
accidental collision than hashes.

> In fact, in this discussion about readable hash codes, we're reinventing
> variations on this kind of mechanism.
> 
> While I do use hash codes for speed in many situations, I always back them
> up with another. theoretically correct mechanism.

Anyway, the really interesting use of hash codes is in the
communications protocol, where they are irreplaceable -- there isn't a
another theoretically correct mechanism.  (Except, "go really really
slow".)

Just "fixing" the revision ids alone won't help (much); most hashes
generated and compared when using monotone are in the networking
layer...

-- Nathaniel

-- 
Details are all that matters; God dwells there, and you never get to
see Him if you don't struggle to get them right. -- Stephen Jay Gould