>If you use a 'population' of searches, combined with multiplication
>(split the ones at a good point in space and remove the bottom
>feeders) and random bitflipping (flip a random bit in 10% of the
>searchers) you might be able to get around the problem.
>
>Rik
-matt