Re: [tlug] Are ordered hashes useful?

["Stephen J. Turnbull" <stephen@example.com>]

Marty Pauley writes:
 > On Sat, Feb 12, 2011 at 1:47 AM, Stephen J. Turnbull <stephen@example.com> wrote:

 > > But a hash can't do that, it's a deterministic algorithm and therefore
 > > always gives the same answer from the same inputs.
 > 
 > But then it's easy to turn that into an algorithm that can be
 > different between program runs by choosing a random number at the
 > start of the program and modifying (or "salting") the key with it
 > before hashing.  If this happens automatically then from the users
 > point of view it is part of the hashing algorithm.

It's rarely a good idea to assume that users have a clue.  This is not
an exceptional case. :-)  Technically, that is not part of the hash
algorithm.  Since we were talking about implementation of various
semantics via hash algorithms, the technical point of view is the
appropriate one.

What is happening under the hood here is that the data, not the
algorithm, is being modified before the hash is applied.  And the salt
must be saved so that the algorithm can be repeated.  If it can't, what
good is it?