"Mark Mitchell" <mmitchell@riccagroup.com> writes:
> I think it's very obvious that Postgres developers have no interest in
> going over 1600 columns in the foreseeable future and which forces us
> to find creative ways around it but I just don't see why it has to be
> this way.
Well, it's a tradeoff. Supporting > 1600 columns would require widening
t_hoff, which means another byte occupied by row headers, which is a
data structure that we have sweated blood to minimize and aren't eager
to bloat just to support what seems extremely dubious database design
practice. The other possible inefficiencies are minor by comparison
to that objection: larger row headers are a cost that will be paid by
*every* user of Postgres.
regards, tom lane