On Jan 5, 2004, at 1:38 PM, Bruno Wolff III wrote:
> I think this is probably the issue with foreign key checks needing an
> exclusive lock, since there is no shared lock that will prevent
> deletes.
>
That was my original thought upon reading all the discussion of late
regarding the FK checking locks. I figured if I deferred the checks to
commit, I could save some contention time. However, if FK checks are
skipped if the field in question is not updated, what locks would there
be? Are they taken even if the checks are not performed on some sort
of "be prepared" principle?
Vivek Khera, Ph.D.
+1-301-869-4449 x806
