Bruce Momjian <pgman@candle.pha.pa.us> writes:
> My basic idea was using INFO-like message to send the SET change
> information.
Well, that's not really materially different from what I was proposing,
except that I was thinking of a format in which the variable name and
value are separate fields of a protocol message. If we send it out
as an INFO-message-like thingy then the client library has to parse
that string...
regards, tom lane