On Mon, Oct 9, 2023 at 11:09 AM Jelte Fennema <postgres@jeltef.nl> wrote: > Since the protocol already returns OIDs in the ParameterDescription > and RowDescription messages I don't see why using OIDs for this GUC > would cause any additional problems.
...but then...
> With Citus the same user defined type can have > different OIDs on each of the servers in the cluster.
I realize that your intention here may be to say that this is not an *additional* problem but one we have already. But it seems like one that we ought to be trying to solve, rather than propagating a problematic solution into more places.
Decisions we make about the wire protocol are some of the most long-lasting and painful decisions we make, right up there with the on-disk format. Maybe worse, in some ways.
So if we use <schema>.<type> would it be possible to have something like <builtin> which represents a set of well known types?
My goal here is to reduce the overhead of naming all the types the client wants in binary. The list of well known types is pretty long.
Additionally we could have a shorthand for removing a well known type.
