Re: ACLs versus ALTER OWNER

Christopher Kings-Lynne <chriskl@familyhealth.com.au> writes:
>> Someone else suggested having pg_dump dump all objects without ownership
>> (so, on restore, they'd all initially be owned by the user running the
>> script, hopefully a superuser) and then doing ALTER OWNERs and GRANTs at
>> the bottom.

> Actually, this would probably only be reasonable if you fixed the ACLs 
> after an ALTER OWNER, like you proposed earlier.

I was envisioning pg_dump not issuing any GRANTs until after the
ALTER OWNER steps, so it really wouldn't matter whether ALTER OWNER did
anything to the ACL list; it'd still be NULL at that point anyway.
(I do, however, have every intention of fixing ALTER OWNER that way
before 7.5 freeze.)

BTW, is pg_dump careful about the order in which it issues GRANTs?
Specifically, what about being sure that chains of GRANT OPTIONs
are re-granted in a legal sequence?  I don't recall any smarts in
the code about that...
        regards, tom lane