Re: Unstoppable spam

   From: Tracy Reed <[email protected]>

   Ah. I had forgotten that I got crmlearn from:

   http://crm114.sourceforge.net/docs/CRM114_Mailfilter_HOWTO.txt

   which presents the script as such:

      grep -a -v "^X-CRM114" | \
       /usr/share/crm114/mailfilter.crm -u $HOME/.crm114/ $1 | \
      grep -a "^X-CRM114"

   for piping into mutt using this in the .muttrc:

   macro index \es "<pipe-entry>crmlearn --learnspam\n<save-entry>=3Dspam/done=
   \n" \"crm114 learn as spam, save in spam/done"

   which is how I have things setup.

Ahhh... so I did write it, once a long time ago.  Sorry.

   > Without seeing the "crmlearn" program, it's hard to guess what
   > might be going wrong.
   >=20
   > Do you get any error messages when you do "crm --learnspam" ?

   No, no error messages. crm114 does emit a success message of some sort
   but it flies by so fast I don't usually see it before mutt redraws the
   screen.

   > Can you learn anything *else* and have it take effect?

   Yes. Lots of other things do learn and take effect. And I notice that
   it appears to have started catching some (I would guess almost half)
   of the spams of this form. But half still make it into my inbox
   also.

Hmmm... so the real problem is that the particular spammer is
really good at leaving very few traces for the statistical analysis
to get a model of.

But- with enough samples, a pattern starts to emerge- and now that 
pattern is enough to get half of these spams.  I'd say to keep
training then; the hill-climb learning is getting a target lock.

	 - Bill Yerazunis

------------------------------------------------------------------------------
Enter the BlackBerry Developer Challenge  
This is your chance to win up to $100,000 in prizes! For a limited time, 
vendors submitting new applications to BlackBerry App World(TM) will have 
the opportunity to enter the BlackBerry Developer Challenge. See full prize 
details at: http://p.sf.net/sfu/blackberry