Jonathon Delacour the heart of things

I'm not sure these programs work with Mac's, but I'm more than ready to take some draconian actions.

Apparently commenting on others' web pages makes one a target for spam.

I'm so upset with the constant porn that I receive that I'm not sure anything could be overkill, Jonathon.

SpamAssassin doesn't actually delete any mail. It assigns a score to each mail, and if the score is higher than the threshold you set, it adds a *****SPAM***** to the subject. What you then do to the mail is up to you.

My experience is that it misses about 2-5% of the spam, and marks about 1 in 1000 real mails as a false positive.

I have used mailshell.com's free disposable address/redirection service (though I'm not sure if it's free anymore) for a couple of years now. Superb.

Loren, OS X Mail's built in Junk filter does very well for me.

A very few false positives (eg spam parodies), it misses < 1% (generally the short ones that are composed to sound like a friend's message with no upsell and a single link in). Also, it has a good nose for bad ad copy, even from sites I did sign up for, and their boring offers get junked too.

It has an implicit whitelist too, in that anyone who is in your address book, or you have sent a reply to is whitelisted.

Killing email from hotmail.com, yahoo.com, and aol.com will probably cut your spam in half.

I had a client who used a hotmail address for a trip to Japan and she found she couldn't download the genuine emails during the limited time she had to access the web. She is now very happy using Neomail while travellng for accessing her normal mail.

On a Mac I find SpamSieve absolutely brilliant. It uses Bayesian filtering and learns as you go. Apple's Mail does the same. [Is that 12" powerbook getting more tempting?]

SpamAssassin does a really good job, and I've had a single instance of a legitimate e-mail getting blocked. It happened to be from someone in China, and it flagged it because of a series of multibyte entities strung together throughout the message--these were characters, of course, but it didn't know that.

Thankfully I was able to simply able to tell it never to block messages from that address.

Spamassassin is very configurable: you can teach it new rules, add whitelists/blacklists, assign different weights to existing rules (each rule-match adds N points to a message's spamminess-score), and set the threshold for marking something as spam.

I've had good luck with it, with very little tinkering. The only false-positives I've observed have been from senders with compromised ISPs.

It's interesting to observe the emergence of spam that's been carefully worded to "slip under the radar."

Apple's Mail deos not use Bayesian filtering, it uses Latent Semantic Mapping. Tim Oren explains the different here:
http://www.pacificavc.com/blog/2003/02/10.html#a78

And I don't want 'configurable' spam filters. The last thing I need is to spend my time configuring a spam filter.

I don't know about Yahoo, but from my experience, killing Hotmail accounts would be overkill: I have a fair number of not-very-computer-literate and/or non-computer-owning friends who use Hotmail because that's all they can manage.

The system my ISP uses - I don't actually know what it is, but it may be SpamAssassin - flags all suspected spams with '|-_-|', so I've set up a filter to throw those into the Trash folder (in Eudora). It correctly flags no more than about 80% of spam, with rather too many - perhaps 1-2% - of false positives.

Dunno if that helps, but it's there anyway!

Loren, I've stopped including my real email address when I leave comments on other people's sites. I followed Mark Pilgrim's example (a@b.com) and enter my address as j@d.net.

Scott, as soon as I read your comment that Spam Assassin doesn't delete any mail I switched it on.

Stav, I think you might have mentioned mailshell.com before. I'm not sure that I'd get the value from it, given my SpamAssassin/SpamKiller combo and the fact that my hosting service provides NeoMail for when I'm away from my desktop or notebook machines.

Anyway, thanks for all the other feedback. And for the Macintosh advice (which I may need within a month or two).