Re: OID wraparound (was Re: pg_depend)

Lamar Owen <lamar(dot)owen(at)wgcr(dot)org> writes:
> On Wednesday 18 July 2001 16:06, Tom Lane wrote:
>> It remains to be debated exactly how users should control the choice for
>> user tables, and which choice ought to be the default. I don't have a
>> strong opinion about that either way, and am prepared to hear
>> suggestions.

> SET OIDGEN boolean for database-wide default policy.
> CREATE TABLE WITH OIDS for individual tables? CREATE TABLE WITHOUT OIDS?

Something along that line, probably.

> ?? Is this sort of thing addressed by any SQL standard (Thomas?)?

OIDs aren't standard, so the standards are hardly likely to help us
decide how they should work.

I think the really critical choice here is how much backwards
compatibility we want to keep. The most backwards-compatible way,
obviously, is OIDs on by default and things work exactly as they
do now. But if we were willing to bend things a little then some
interesting possibilities open up. One thing I've been wondering
about is whether an explicit WITH OIDS spec ought to cause automatic
creation of a unique index on OID for that table. ISTM that any
application that wants OIDs at all would want such an index...

regards, tom lane