Hi, from your description it seems that an "I have seen this before" component would do well. The IXHASH was originally developed for just that context: if the same mail is sent to almost everybody in a <50 usergroup, the recipients are likely not to want it. Consideration (if you want to handle the load ... you probably can): your MTA should reject unknown recipients, but it could also reject after the data phase and immediately feed to bayes
Wolfgang Hamann
