Hi Mike,
>Good. Now speculation: If a system was built with VFP cursors in our "simulated" / not physically separate COM ntier approaches, they will be faster also?
>
>Has anyone done any benchmarking of a query of differing complexities from SQL via COM and via VFP Cursors?

I don't think such a benchmark could ever be "fair", as there probably are different timing differences depending on reccount(), recsize() of the cursors and so on.
When doing layers I think the most important rule is to define "chunky, not chatty" interfaces - then marshalling will hurt less if it has to occur.

my 0.02 EUR

thomas