> On Oct 7, 2021, at 7:44 PM, Stephen Frost <sfr...@snowman.net> wrote:
>
> I don't actually think REVOKE ROLE CASCADE must not fail, nor do I see
> that as explicit in anything you quote above.
I don't see that myself, but I thought that you would, given your other
statements about how we shouldn't take a spec requirement to do X and turn it
into doing X+Y, because the user wouldn't be expecting Y. So I thought that if
DROP ROLE bob was defined in the spec to basically just do REVOKE bob FROM
EVERYBODY, and if the CASCADE version of that wasn't supposed to fail, then
you'd say that DROP ROLE bob CASCADE wasn't supposed to fail either. (Failing
is the unexpected action Y that I expected your rule to prohibit.)
—
Mark Dilger
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
