Jason -

In further testing I'm finding that the performance
differences primarily hinge on how the 'LIKE' predicate is
used.

If I use a non-indexed column it is frightfully slow,
which is what you would expect. If I use an indexed
column, the performance varies significantly based on the
size of the result set.

On Windows I can return result sets of almost any size
very quickly, which would seem to indicate that once the
server determines the result set, the components use
caching somewhere to speed up the process and return only
a portion of a large result set.

In looking at the TIBODatabase component I don't see
anything significant as far as properties as they relate
to caching.

Regardless, all settings on both the TIBODatabase and
TIBOQuery are identical.

Obviously the processor is used differently in Linux v
Windows. In watching the gnome monitor the processor
continually cycles regardless of what operation is
ongoing. In short it does not appear that the kylix/ibo
app is generating any noticable overhead, which would seem
to indicate that the bulk of the processing is on the
database server where it should be.

FYI, I"m running the Windows query in a VMWare session
running W2KPro, so I'm literally copying and pasting sql
statements from host to host.

Anything obvious jump out at anyone?