Why Bayesian spam-filters are good
Learning from experience means they don't make the same mistakes twice.
From today’s NTK:
If SAM KINGTON can read this at all, it’s probably via our RSS feed, after his SpamSieve installation achieved an superhuman level of Bayesian sentience and rejected NTK 2003-08-08 for its gratuitous use of such giveaways as “organised”, “mitnick”, “mercy”, “preceded”, “erotic”, “non-” and (apparently) “3210”.
I’d like to point out that, after I told SpamSieve that that email from NTK was not, in fact, spam, it hasn’t rejected any further emails from them since. (In fact, tips@spesh.com has been automatically whitelisted.)
The only thing I have to do these days with spam is to go through my Spam folder every day eyeballing for false positives. I get maybe one or two a week, it takes something like a minute to go through a day’s spam - seems like a good enough solution to me.
Let me say this, because we hear so many tales of woe about how bad the spam situation is and how we need to throw away email as it stands and reinvent it, possibly using micropayments or signing or massive whitelists or something.
Email is not broken. We are winning the fight against spam.
(As an aside: yes, I was tickled pink by getting mentioned in NTK :-) )