-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Mon, Sep 22, 2003 at 01:23:26AM +0100, Karsten M. Self wrote: > 2. Content-based filters can't distinguish SPAM from legitimate mail > with sufficient accuracy. > >What is the supporting basis for this statement?
I think the word "sufficient" makes this statement subjective. As I see it, there are three variables: 1. Quantity of incoming spam. 2. Accuracy of the content-based filter. 3. The individual's acceptible level of spam in the inbox. Until the accuracy of the filter reaches 100%, you can always ratchet up the quantity of incoming spam or reduce the user's spam tolerance for the filters to be insufficient. Personally, filters worked fine for me until my incoming spam level reached about 50 per day. Then there were too many spams getting to my inbox for me to consider the filter "sufficient." If your filters find spam 99% of the time, someone who can't stand spam on a daily basis will still be unhappy when the spam level hits 100 per day. Note that when I switched to TMDA I observed that the spam I'd received in the previous eight months was equal to all the spam I'd ever received in the six years before that, so I could expect the problem to get worse. I saw a pro-CR article about a guy who got 3000 spams a day. Even with 99% true positives, 30 spams a day is a mess that few would find acceptible. On the other side, if someone only has three spams a day incoming, virtually any filter will be good enough. - -- Kyle Hasselbacher | Why is it that we rejoice at a birth and grieve at a [EMAIL PROTECTED] | funeral? It is because we are not the person involved. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (GNU/Linux) iD8DBQE/bv4T10sofiqUxIQRAsSDAKCR6TZiHMhrDKL6R4vhQ1WqO8dk8gCeJU3q DH+1NIIyNrhRnspfM5hqCmM= =T4CT -----END PGP SIGNATURE----- _____________________________________________ tmda-users mailing list ([EMAIL PROTECTED]) http://tmda.net/lists/listinfo/tmda-users
