on 5/30/02 2:05 PM, Busby@xxxxxxxx purportedly said:
> Dear List,
> Thank you for the prompt responses...I read the MAN on OID and also
> on SERIAL...the benefit of using the MS-UUID is that the identifier created
> is guarenteed to be unique in spacetime. OID and SERIAL are not...they are
> only guarenteed to be unique in a database.
>
> Perhaps I should have mentioned that I'm building a multi-database solution
> (20+ databases) in which all 20+ DBs must use the same identifier across
> databases for some objects (ex: Automobile Brands) but their own identifier
> for their own data (ex: Accouts/Clients) this way when the child database
> publish to the master there is no possiblity of some object having the same
> identifer as another...and the object identifier can stay the same across
> all 20+ DBs.
>
> The GUID from Micro$oft is formatted like
> {01234567-89AB-CDEF-0123-456789ABCDEF} it represents a 16 byte number that
> is again unique in spacetime. Is there PostgreSQL solution for something
> like that or will I have to come up with my own.
AFAIK Postgres doesn't have such a feature built-in, so you would have to do
it yourself. If Db access is entirely web-based, and you use Apache,
mod_unique may do what you want, and it is supposed to be unique throughout
time. However, it may not be unique across multiple Apache installations, if
you have such an environment.
To use Postgres alone, you would likely need a particular database that
"tracks" unique ids. It adds a bit of overhead, and you face parallel access
issues. You can mimic a 16-byte number using 4 INT4 fields (unless your
platform supports 64 bit numbers), but you would have to watch for overflow,
which can get confusing since Postgres doesn't have the notion of unsigned
integers. Fortunately, you can set up triggers/stored procedures that will
handle all of this automatically.
Keary Suska
Esoteritech, Inc.
"Leveraging Open Source for a better Internet"
