That's cleared that up, thanks!
Tamsin
-Original Message-
From: Jan Wieck [mailto:[EMAIL PROTECTED]]
Sent: 04 September 2000 15:50
To: Tamsin
Cc: [EMAIL PROTECTED]
Subject: Re: [GENERAL] permissions & foreign keys
Tamsin wrote:
>
> I don't really see why it wants to update feedback_type? Can anyone tell
me
> what I'm doing wrong, or will I just have to grant update on feedback_type
> (and all other tables referenced by FKs)?
>
It doesn't want to update it. It just does the SELECT ... FOR
UPDATE to lock the now referenced row. Doing it without a
lock would make it possible, that just after your backend
checked that the PK row exists but before you got a chance to
commit, another backend could delete that PK without seeing
your just inserted reference. End would be a violated FK
constraint.
The bug here is, that doing a SELECT ... FOR UPDATE already
requires UPDATE permissions. The correct solution would be to
require a REFERENCES privilege for the owner of the
referencing table. But we don't have that up to now.
Maybe I can do something about it for 7.1.
Jan
--
#==#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#== [EMAIL PROTECTED] #
