On Wed, Aug 14, 2002 at 01:56:57PM -0700, Jon Swinth wrote:
> ah, now I understand where you got FK from. The transaction exeception I run
> into most often is caused by an Unique Key (not the PK either). An insert
> will block on UK violation when the existing record has been inserted from a
> non-complete transaction.
Ok, so these are just separate issues. Sorry, I'm especially dim
this week (we're moving offices is my best excuse).
> As for your pending and posted idea, are you proposing to not have FK on the
> pending table? What do I do when the order fails an FK when moving from
> pending to posted? The whole point of the transaction is that when I am
> done, everything is updated properly or nothing is updated.
No, you should have the FKs on the pending table. Hmm. I see, now:
the problem may be related also to the long-running transaction,
because you end up having to take the lock for the duration. So
never mind all of what I said.
> Based on what I know of Postgre so far, there are two ways to solve the FK
> lock issues. Both require that the concept of read lock be added to the core
> of postgre.
Yes, I think this is right. And yes, that lock mechanism would be
valuable.
A
--
----
Andrew Sullivan 87 Mowat Avenue
Liberty RMS Toronto, Ontario Canada
<andrew@libertyrms.info> M6K 3E3
+1 416 646 3304 x110