> "Henryk Szal" <szal@doctorq.com.pl> writes:
> > YES, this feature should affect ALL locks.
> > 'Timeout on lock' parameter says to server "I CAN'T WAIT WITH THIS
> > TRANSACTION TOO LONG BECAUSE OF (ANY) LOCK",
>
> It still seems to me that what such an application wants is not a lock
> timeout at all, but an overall limit on the total elapsed time for the
> query. If you can't afford to wait to get a lock, why is it OK to wait
> (perhaps much longer) for I/O or computation?
Yes, that is a valid argument. The only thing I can counter is that (in OLTP)
it is usually easy to predict the amount of work that needs to be done
for your own tx (we are typically talking about 1 - 200 ms here), but it is not easy
to predict how long another session needs to complete it's transaction
(the other session might be OLAP, vacuum ...).
Andreas
