Re: [Monotone-devel] Re: RFC: Fake IDs

[Daniel Carosone]

On Wed, Jul 19, 2006 at 12:04:55AM -0700, Graydon Hoare wrote:
> My concern with your patch is that all it guarantees is that two 
> different lines of code don't produce the same hash. You are then 
> obliged to confirm that no two *dynamic control paths* use the same line 
> of code to produce fakes that are used in ways that conflict.

This was my concern as well.  Standard 'cheap tricks' like throwing
the current time, stack addresses/contents, etc into the hash might
suffice at least to get us back to the types of probabilities we deal
with for real revisions.

Or there's an alternate approach. If the concern is that two fake id's
clash with eachother, rather than that either of them might clash with
the db, we can keep the history of fake id's generated in memory.
From there:
  - we either explicitly detect duplicate fakes, and perhaps learn
    something interesting in the process, or
  - we throw the history-of-fakes into the hash, and get the same kind
    of protection against dupes as we get for real revisions.

--
Dan.

pgpYLv4I_U7WG.pgp
Description: PGP signature