Comment 23 for bug 9870

> why we don't just find a tool which uses RBL's and be done with the whole thing?

Because self-learning filters generally perform a lot better after being trained, when considering both false negatives (no problem if there aren't many) and false positives (even 1 might get you in trouble). As (almost) every RBL I know causes lots of false positives, I don't think they are a good idea.

Also, it would make it possible for a third party to hurt Ubuntu users badly (those RBLs are not under our control).