"Life is hard, and then you die" <[email protected]> writes:

> Thanks for your answer. But since the isolation-level is read-uncommitted
> I don't see why it needs to lock the row unless it is actually going
> to update it. Just reading the row does not require a lock, as the
> following query will run quite happily just before the update:
>
>     con1: SELECT * FROM TEST WHERE version = 1

Right, but I think that since this is an update operation, it will try
to get an (exclusive) lock right away (otherwise the row might change
between the point where we qualify it and try to update it)..

Dag

Reply via email to