Hi Jim, Hi Richard, thanks for your fast replies. You both mention that it is nearly impossible to reach 100% correct identification of human writing as administrative vs. 'valid' contribution. Agreed. 100% will be impossible (how about training a neural net ? -- na, I'm just kiddin' :-). But how about making a decision lets say of 90% correctness and put the other 10% on hold until the listowner/editor decides (like Send= 'Tested')? The reason for asking this is, I've proposed a gateway between a BitNet-list and its UseNet counterpart. One objection (so far the only) of the UseNet side was that they didn't like to see the 'unavoidable-unsubscribe-me messages'. Although I agree that educating the subscribers is THE solution, the subject of the discussion (the programming language SMALLTALK) has a tendency to be misinterpreted by people who want to chat, and are novices who obviously do not read the blurp that is sent out to each subscriber. -joerg