Csaba Nagy <nagy@ecircle-ag.com> writes:
> Actually, if there were a working and maintained pg_upgrade, I'm pretty
> sure nobody would use pg_dump as an upgrade facility anymore.
You think so eh? Hint: the only workable design I've seen for pg_upgrade
uses pg_dump as a component. It's much easier to handle
version-to-version changes in pg_dump than it would be inside the server.
Example: there is no way that a pre-8.1 server could be expected to know
that it had better set standard_conforming_strings = off to ensure that
the SQL it's emitting will be understood properly by a post-8.3 server.
regards, tom lane