On Wed, 11 Dec 2002 the voices made Justin Mason write:

>   (a) it lets legit publishers avoid relatively-obvious trouble areas
>   (like talking about spam laws etc.)

 Like everyone already could do, but only spammers bothered to do.

>   (b) it lures spammers into a false sense of security, as once they send
>   their spam the massive quantity of forged headers will give them away
>   anyway ;)

 Not to mention DCC etc.

>   (c) the Bayesian stuff will soon muddy the waters quite a bit more.

 Personally I found that I wasted too much time worrying about FPs when using
SA; and I didn't like the amount of time I put into mailfiltering. So I stopped
using SA, and let my procmailrc-spam file grow instead; I found that I could
get a much better result with my own rules than I could ever get with SA.

 Then I tried a Bayesian solution (just picked the first one that looked good,
which happened to be Bogofilter), and found that even without a perfectly clean
spam/ham-collection (I was lazy, and used two "good enough" mailboxes just for
testing) I got the "WOW, this works better than I thought"-experience.

 The results looked good; I removed my own spamfilters, moved Bogofilter to the
top of my procmailrc (ie started filtering even mailinglists that are spamless,
mostly spamless and lists that I'd whitelisted because of the spam-related
talking going on) and let it learn from whatever it classified as spam/ham.


 Instead of letting it "guess" even though there are some sure signs of spam
I'm going to add some of my own "rules" again (mostly network/header-related
stuff), which should reduce the very few FNs I've gotten...

 I can't help but think that a perfect solution would be to let SA do network
and header tests, add the results as separate headers (x-sa-[name_of_test]) and
then let each of those headers be a token in the Bayesian filtering; ie
removing all other rules.


        /Tony
-- 
# Per scientiam ad libertatem! // Through knowledge towards freedom! #
# Genom kunskap mot frihet! =*= (c) 1999-2002 [EMAIL PROTECTED] =*= #

     perl -e'print$_{$_} for sort%_=`lynx -dump svanstrom.com/t`'



-------------------------------------------------------
This sf.net email is sponsored by:
With Great Power, Comes Great Responsibility 
Learn to use your power at OSDN's High Performance Computing Channel
http://hpc.devchannel.org/
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to