On Wed, Aug 09, 2006 at 03:16:29PM -0400, Bill Moran wrote:
> In response to Matthew Seaman <firstname.lastname@example.org>:
> > This problem is intractable: any scheme you can think of to generate a
> > unique identifying number on a random host out there on the net will either
> > fail to actually be unique, or suffer from mutating over time as machine
> > configuration changes.
> Really? What if you just generate some sort of UID or GUID and store it
> in /var/db/bsdstats.guid (or similar)?
Well, exactly. What I neglected to say in the above was "to generate
a unique identifying number that encodes part of the machine
However, you're right in that the client could just invent its own
random ID number. Given the large number of possible ID numbers in
the scheme I proposed, there shouldn't be any problem with collisions
so long as all those machines are generating good random numbers.
On reflection, the advantages of having the server generate the ID
numbers are not really all that compelling.
 In fact, it would be a pretty neat experiment to get a whole load
of machines to generate a chunk'o'randomness and send it into a central
machine and see just how evenly distributed the answers are.
Dr Matthew J Seaman MA, D.Phil. 7 Priory Courtyard
PGP: http://www.infracaninophile.co.uk/pgpkey Ramsgate
Kent, CT11 9PW