> I have seen nicer fast random row implement, but that will work.


Do you happen to have a snip of it, the one I have seems to lean
pretty heavy as far as I can tell, and on occasion, though rare, also
sends me an empty result set.
--
Scott
talklists@newgeo.com