> So far I haven't thought of a better idea, I'm afraid. Maybe there isn't one. How about the approach Nathan described in his 1997 paper: http://www.cantrip.org/emptyopt.html Felix