[Mailman-Developers] Re: Anti-spam "killer app"?

[David Champion]

* On 2002.08.22, in <3D651472.5B1FB99D@nleaudio.com>,
*	"Bob Puff@NLE" <bob@nleaudio.com> wrote:
> My biggest question to all this is: how do you add to the good list,
> and the bad list? It seems like you need to moderate each message, at
> least in the beginning.

Yes, but I find it doesn't take any longer than than deletion. I only
need to react when the filter misidentifies something. I can easily
bind a macro keystroke to some sequence of feeding it into my database
and deleting it from my mailbox. It's not really any different from
any other spam detection mechanism. Even if there's a little more
startup cost, I expect that over time, maintenance cost is much lower
than keeping my other spam filter rule sets up to date -- and so far,
the false positive rate looks astoundingly good for remarkably little
investment time in the valuation of different canonical properties of
spam ("Nigerian Foreign Minister", "teen delight", "Pay your mortgage",
"add.*inches").

>                         How would this be implemented for standard
> mail clients? Would you forward each good and bad message to special
> addresses? It would seem this would take more time than simply
> deleting the spam.

For the generic mail client, yes; that's what I was thinking of
providing to my site. For specific cases, I can set up macros for UNIX
mail client users, and perhaps install pretty buttons for our webmail
users. But "forward to spam-submit@uchicago.edu" is pretty simple,
and as the filter gets smarter, it becomes less necessary to interact
with it. I've been using mine since Monday, and I'm getting no false
positives anymore. Not much of a sample size, but if I can keep a rate
anywhere near that, I'm not opposed to checking my separate spam folder
every couple of days.


I find that the biggest payoff is not in having a smart agent eliminate
my spam (I don't trust computers enough to allow that by any algorithm),
but in helping me to think about messages correctly. Sometimes my
mind is just not quite in gear, and I need to read a bit of something
before I realize that it's spam. If an agent can set a flag and thereby
predispose my brain against it -- if I can automatically flag all the
suspected spam and leave it to deal with in one big sweep at the end of
my mail-reading session -- then I'm way ahead of where I was last week
just in terms of brain efficiency.

-- 
 -D.			We establised a fine coffee. What everybody can say
 Sun Project, APC/UCCO	TASTY! It's fresh, so-mild, with some special coffee's
 University of Chicago	bitter and sourtaste. "LET'S HAVE SUCH A COFFEE! NOW!"
 dgc@uchicago.edu	Please love CAFE MIAMI. Many thanks.