> > Thanks a lot for your explanation. But it seems to me, that in situations
> > when tables contain large character fields, buffer I/O will be the most
> > part of a query execution. This will be extremely sad when most of those
> > character fields are nulls or empty.
>
> Right. It would not involve a great deal more processing to
> compress the intermediate output and address it through an
> array... Part of the answer is that returning lots of long
> fields, mostly empty, wasn't part of the design center when
> that code was written.