Josh Berkus <josh@agliodbs.com> writes:
> Yeah, the problem is that with the upcoming "group ownership" I see
> user-based quotas as being rather difficult to implement unambiguously.
> Even more so when we get "local users" in the future. So I'd only want
> to do it if there was a real-world use case that tablespace quotas
> wouldn't satisfy.
There's also the point that having both user- and tablespace-related
limits would mean (at least) double the implementation overhead, for
a lot less than double the usefulness.
I'm with Josh on this one: I want to see something a lot more convincing
than "it would be nice" or "Oracle has it" before buying into more than
one type of quota.
BTW, I think it is actually impossible to do global per-user limits
within anything approaching the current system structure, because you'd
have no way to know which tables of other databases belong to which
user. Per-tablespace quotas can at least be done by reference to just
the filesystem, without needing inaccessible catalogs of other
databases.
regards, tom lane