At 23:23 13.11.2002 +1100, you wrote:

>Oh.. more importantly, though, why was a read-only transaction in a report
>(i.e. queries which was select only - no updates) causing a deadlock in my
>main updating transaction. Even if it was left open after the report
>generated, surely the read-only status would not cause a deadlock?

I have no idea. Any pecularities, e.g. tiConsistency isolation or
PessimisticLocking set to true (I doubt these would cause any such
behaviour, but am not 100% sure)? None of your report do any updating like
setting a field meaning 'This record has now been reported' or 'Last accessed'?

>Wouldn't the amount of work (rollback Vs commit) have something to do with
>updating/versioning which had taken place? so a readonly rollback would be
>no different to a read-only commit?
>It's interesting to get to the bottom of it.

Agreed, but unfortunately nothing I am capable of answering. I think
someone knowledgeable (Ann or similar) recommended using commit a while
ago, but could not find anything looking through the messages I have kept.
In theory I would say that a readonly transaction ought to be able to
figure out that it didn't have to do anything special and treat commit and
rollback the same way, but I don't think it is that simple.

Set