> I certainly think that if you could combine the 'best' bits of both
> architectures into one version, that would be the way to go.

I think (others may confirm or deny) that this is what
we are trying to answer. Can we accomplish this or does
the Classic need to continue on.

> In development it is a must.

I agree.

> In production, it is very useful not only for those bad
> queries you have missed :-), but for 'users' who disconnect
> badly (using the on/off switch).

If using TCP, this is already implemented (as of v5.6 maybe
v5.5 -- I forget -- Charlie??) in SuperServer. The Crystal
reports / adhoc query issue is a killer. I think that improving
our join performance could likely lessen the impact of many
of those CR generated queries.

> You could implement some extra triggers for 'users', AfterLogin,
> BeforeLogout, AfterCrash, AfterTimeOut.

I'm not sure that these would solve this problem but they are
interesting to think about. Many things could be handled
already with the Events in IBX/IBO. The TIBDatabase has
methods for

IdleTimer -- Monitors a time interval for a connection doing
nothing.

OnLogin
Before/After | Connect/Disconnect

> I did not think Superserver scaled at all on SMP machines, I
> thought it took a performance hit?

I suspect that if a performance hit is seen, its likely
due to the OS' processor scheduling mechanisms. I've only
heard troubles with NT and SMP with IB SS, thus IB_Affinity
exists.

Markus