On Mon, Jul 10, 2006 at 03:09:30AM +0200, Anders Breindahl wrote: > That's reason enough for me (albeit, I wouldn't assume that an employer > would be as easily convinced) to continue the non-binary choices of > bayesian and otherwise content-oriented filtering methods, while > something fundamentally fixing the feasibility of spamming is made up. That is how I filter spam, using spambayes, and I must say it is very effective once it is trained. It also has certain distinct advantages as a "global" solution to spam filtering: - All false positives are held on the recipient's box. They are not lost, they do not disappear into a black hole, and the recipient can manually check every few days to make sure the spam filter hasn't caught anything it shouldn't have. - The processing is carried out on the recipient's box. The majority of PCs have huge amounts of processing power which spends most of its time waiting for the user to press the next key. A small amount of this spare CPU time can easily be used for content-based spam filtering without the problems of running a CPU-intensive spam filter on an ISP's mailserver. - The filtering criteria are determined according to what the recipient does and doesn't consider to be spam, not what the ISP more-or-less arbitrarily decides on the user's behalf. (Doctors and people working in the pharmaceutical industry might not appreciate having emails mentioning a certain drug beginning with "V" automatically canned, for example...) It does have the disadvantage that it requires a certain amount of clue on the part of the user, but I suspect this problem can be minimised by an intelligently-designed and friendly user interface. Or am I overly optimistic? :-) -- Pigeon Be kind to pigeons Get my GPG key here: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x21C61F7F
Attachment:
signature.asc
Description: Digital signature