Re: Role incompatibilities

* Peter Eisentraut (peter_e(at)gmx(dot)net) wrote:
> Stephen Frost wrote:
> > Is there a particular issue/problem you're running into? It might
> > make more sense to focus on what you actually need than what the spec
> > says you need...
>
> The particular issue I'm running into is that I'm trying to get the
> information schema up to speed but the current role implementation
> doesn't really match anywhere. I remember the discussion about the
> inherit flag vaguely, and I think I might even have contributed to the
> confusion, but the fact is that some concepts like the purpose of SET
> ROLE and the difference between enabled and applicable roles have
> apparently been misunderstood.

Well.. Applicable roles are roles which you can "SET ROLE" to, but
which you don't automatically get the permissions of (inherit). As I
recall, the spec wants all roles to be like this until an explicit "SET
ROLE" is done. When a "SET ROLE" is done, then that role (and all other
roles granted to it) are "enabled".

In Postgres terms, the "pg_has_role()" function can provide the answer
to both questions, based on what's passed in.

For 'enabled' roles:

pg_has_role('abc','USAGE');

For 'applicable' roles:

pg_has_role('abc','MEMBER');

Where the current user is asking the question "do I have USAGE/MEMBER
(enabled/applicable) rights on role 'abc'?"

At least, I'm pretty sure that's the idea. Hopefully that helps clear
up what should be done in information_schema...

Thanks,

Stephen