[Spambayes] spambayes db

Dhaval Patel dhaval at patel.sh
Sat Jun 10 19:58:10 CEST 2006


Hello all, I am new to the list. I have used spambayes for a long time and have been
very happy with the results. I originally ran it on Slackware with Sendmail/Procmail on
mbox style folders. In this setup each mail user was a real user.

Currently I have changed the setup and now use Debian with Postfix/Maildrop on maildir
style folders. I have virtual users in this setup. I have a vmail user which controls
all the email for the virtual users.

I am still using spambayes for the spam filtering. Even on the new system I have a
seperate Maildrop file for each virtual user and it filters using his own spambayes db file.

When I moved all the users to the new system I converted the mail from mbox -> maildir
and copied the spambayes db file to the new system. When an email comes through it is
being filtered because some spam goes to spam some goes to unsure and a LOT to inbox. :)

I am not able to figure out why so much email comes into the Inbox now. What I have
noticed is that when I run the "sb_mboxtrain.py -d /home/vmail/.spambayes/user.db -g
/home/vmail/user/cur -s /home/vmail/user/.spam/cur" it goes through and reads all the emails

Training ham (/home/vmail/dhaval/cur):
  Reading as MH mailbox
  Trained 666 out of 666 messages
Training spam (/home/vmail/dhaval/.spam/cur):
  Reading as MH mailbox
  Trained 1231 out of 1231 messages

But the problem is that the size of the db file still stays the same (the timestamp is
updated). I have noticed that in the past, it has always become larger. Currently after
training a brand new db file (12288 bytes) with the 666 good messages and 1231 spam
messages, I get a db file that is only 2600960 bytes. 

I though that it was the spambayes package which came with debian so I removed it and
downloaded the package from http://spambayes.sourceforge.net/ and installed it but I
still get the same results.

If anyone has any suggestions on how I can go about troubleshooting this please let me know.


Thanks,
Dhaval


More information about the SpamBayes mailing list