Dumb Bayes question

Gareth list-mailscanner at linguaphone.com
Sun Aug 12 21:11:22 IST 2007

I run spamassassin on my home system which is very low usage and bayes is
working very well.
In fact I would say that bayes works better on small systems for a couple of

1) Bayes works best when it learns your individual mails. For companies
which deal with lots of different topics and areas bayes has to learn a lot
of tokens as being ham. For home servers and small organisations they
receive a far small variety of ham messages so bayes can work better.

2) The default bayes database size is geared towards the smaller user. In my
company if I used the default bayes size (which a lot of people probably do)
the oldest token age would be about 2 days which is far too short for it to
be very effective.
  -----Original Message-----
  From: mailscanner-bounces at lists.mailscanner.info
[mailto:mailscanner-bounces at lists.mailscanner.info]On Behalf Of Steve Mason
  Sent: 12 August 2007 20:35
  To: 'MailScanner discussion'
  Subject: Dumb Bayes question

  Hi all.  I've been running MailScanner for a while now my home server (4
users) and 2 small non-profit organizations, both around 10 users each.
  It seems to me I read a while ago, that Bayes isn't too effective on
low-volume servers due to it not seeing many messages.  Of course I can't
seem to find where I read that now.

  If I use a "starter" database, and occasionally feed it any false
positive/negative messages, is Bayes worth using for very small sites?


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mailscanner.info/pipermail/mailscanner/attachments/20070812/0c139f1a/attachment.html

More information about the MailScanner mailing list