On Wed, 2023-03-22 at 14:42 -0400, Tom Lane wrote:
> This isn't going to help much unless we change the wire protocol
> so that RowDescription messages carry these UUIDs instead of
> (or in addition to?) the OIDs of the column datatypes. While
> that's not completely out of the question, it's a heavy lift
> that will affect multiple layers of client code along with the
> server.
I'm not sure that's a hard requirement. I pointed out a similar
solution for type names here:
https://www.postgresql.org/message-id/4297b9e310172b9a1e6d737e21ad8796d0ab7b03.camel@j-davis.com
In other words: if the Bind message depends on knowing the OID
mappings, that forces an extra round-trip; but if the client doesn't
need the mapping until it receives its first result, then it can use
pipelining to avoid the extra round-trip.
(I haven't actually tried it and I don't know if it's very reasonable
to expect the client to do this.)
> Also, what about container types? I doubt it's sane for
> array-of-foo to have a UUID that's unrelated to the one for foo.
> Composites and ranges would need some intelligence too if we
> don't want them to be unduly complicated to process.
That's a good point. I don't know if that is a major design issue or
not; but it certainly adds complexity to the proposal and/or clients
implementing it.
Regards,
Jeff Davis
