We are using spamassassin at the gateway level with exim. Is it a good idea to use bayes as we don't know which is ham or spam - and the users are unlikely to give us the feed back from different system. In that case bayes learning ability will be compromised.
If bayes can be used how can I modify 23_bayes.cf to give me a lower score to avoid false positives, bearing in mind that sa-updates might overwrite 23_bayes.cf? Regards Sujit Choudhury