> Is this important enough to back-patch? We can't force initdb in back
> branches, but we could suggest that people could drop and re-create the
> information_schema (I think that's supposed to work).
I wouldn't worry about emphasizing the rebuild. Most users don't use
information_schema, and for those who do the refconstraints view is
known to be useless (just try using it for a multicolumn FK).
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com