On Thu, 11 Nov 2004, Terry Lee Tucker wrote:
> Is it ok to put a unique index on the oid for my tables? We are in the process
> of moving from Progress Software to PostgreSQL. In the Progress world, you
> can always uniquely, and quickly find a record by using their version of oid,
> which is recid. I remember reading somewhere that the oid could be
> duplicated across the cluster, but would not be duplicated in a single table.
> Maybe I dreamed it. What is the recommendation regarding this and why?
Well, some system tables that want unique oids already do this. One issue
with making a unique index on oid is that after oid counter rollover
you'll potentially get intermittent constraint violations due to duplicate
oids that you need to be able to handle.