I was thinking that this would help stop OID wrap around while not totally
breaking clients that used OIDs as row identifiers as they'd now have the
int4 primary key value (although I guess there could be risks if the client
assumes there'd be globally unique). Also the primary key would have to be
placed into the OID in all places it could be referenced (for WHERE
clauses,etc...). It'd only work on those tables that had int4 priamary keys,
but I suspect thats a fair few. I don't know wether this'd be worth while,
but was rather throwing it out for thought.
> -----Original Message-----
> From: Hiroshi Inoue [SMTP:Inoue(at)tpf(dot)co(dot)jp]
> Sent: Tuesday, July 24, 2001 2:37 AM
> To: Henshall, Stuart - WCP
> Cc: 'pgsql-hackers(at)postgresql(dot)org'
> Subject: Re: [HACKERS] RE: OID wraparound (was Re: pg_depend)
> "Henshall, Stuart - WCP" wrote:
> > Would it be possible to offer an option for the OID column to get its
> > from an int4 primary key (settable on a per table basis maybe)?
> > - Stuart
> Sorry I don't understand well what you mean.
> What kind of advantages are there if we let OIDs be optional
> and allow such options like you offer ?
> Hiroshi Inoue
pgsql-hackers by date
|Next:||From: tamsin||Date: 2001-07-24 09:15:42|
|Subject: Bad timestamp external representation|
|Previous:||From: Thomas Swan||Date: 2001-07-24 05:54:46|
|Subject: Re: plpgsql.|