[Spambayes] Feature Request

[Bryan D. Andrews]

Don't you agree that you should be able to whitelist everyone you are
doing business though?

Do you not think this would make a great addition?

Thanks.


-----Original Message-----
From: Skip Montanaro [mailto:skip at pobox.com] 
Sent: Monday, April 18, 2005 11:10 AM
To: Bryan D. Andrews
Cc: spambayes at python.org
Subject: Re: [Spambayes] Feature Request


    Bryan> I literally get so much spam now that I cannot even *think*
about
    Bryan> looking in the junk folder -- I just have to trust... the
junk
    Bryan> suspects is getting this way as well...

Hmmm...  How large is your junk suspects folder that you can't at least
skim
it?

Shifting gears a bit, I think it would be worthwhile for all
organizations
to at least look at greylisting as a way to reduce the amount of spam
entering their networks.  Take a look here:

    http://www.greylisting.org/

I happen to use postgrey with postfix as my weapon of choice.  On March
25th
I started using this setup.  This graph shows the number of messages
SpamBayes has had to classify since early last November:

    http://manatee.mojam.com/~skip/sb.png

Note the huge drop in mail it classified as spam after March 25th.  By
my
crude calculation I've seen a drop in spam reaching SpamBayes by a
factor of
four to five, about 1540 per day over the first 20 days of the graph's
period and about 330 per day over the last 20 days.  Note that these
numbers
are for the email received by one person, not an entire organization.
Eliminating the high-scoring spam which is simply discarded (that
scoring >=
0.80), I'm left with about 30 messages classified as spam per day that
need
to be scanned and about twice as many unsures.  Those numbers may just
be
artifacts of the train-to-exhaustion regimen I use or misclassified
mails in
my training database.  At any rate, they are small enough for me to skim
quickly at various times during the day.

Anecdotally, SpamBayes seems to be doing a better job classifying the
mail I
do receive (60 unsures per day notwithstanding).  My theory is that
eliminating entire classes of spam (mostly that sent by
virus-commandeered
Windows machines) probably reduces the variability in the spam that does
get
past postfix, so spam is easier to properly classify.

I realize a postfix mail server is probably not in the cards for
everyone.
If you're a Microsoft customer I suggest you start leaning on them to
add
something similar to their SMTP server product.

-- 
Skip Montanaro
skip at pobox.com