On Sat, 3 Dec 2005, Tom Lane wrote:
> The latter is not really a bug. Languages don't currently have owners
> (ie there is no owner column in pg_language). For ACL-munging purposes
The true behaviour of the ACL may be somewhat unusual - understood (after all,
I admitted at the outset that the use of the "{}" ACLs could seem strange).
> and the privileges aren't going to be enforced against him anyway. So
> the fact that pg_dump doesn't process that part of the ACL isn't very
> meaningful.
True, though that could have unintended side effects in future as certain
details (ACLs or whatever) are lost/modified as a database is dumped/restored
between different PostgreSQL versions (something that was a no-op before
suddenly creates problems a newer version due to feature changes/additions).
Thanks for taking the time to investigate this - my whole concern was just
that the doco-set user expectations be met by the dump tools in reality (ie.,
if a whole database dump is requested, then everything is restored in the
destination cluster).
Thanks again..