Curt Sampson <cjs@cynic.net> writes:
> On Mon, 19 Aug 2002, Jules Alberts wrote:
>> Is there any consensus aboit avoiding OIDs? I'm running a small test
>> system right now using OIDs as a means to refer to BLOBs. Should I
>> expect any trouble using OIDs in our future production system?
> I don't know if there's a consensus, but I certainly avoid using OIDs
> completely in my own tables. They're can wrap, for a start, so in a
> really busy, large database you might end up getting one that you
> already have. Also, I don't like "hidden" fields; if I'm going to refer
> to soemething, I like it to be nice and obvious what's being referred
> to. And of course they're not portable.
As far as using BLOBs goes, you don't have a lot of choice: the lo_xxx
family of functions take and return OID, end of story. Of course
Postgres' notion of a BLOB isn't very portable anyway.
I do agree that for a primary key in a user table, there's no very good
reason to use OIDs rather than using a SERIAL field (ie, a sequence).
regards, tom lane