strange spam mail

[Boris 'pi' Piwinger]

Eric Persson wrote:

> As an example I have put 3 of the mails on the net for you to see, with 
> the resulting x-bogosity line below:
> http://files.persson.tm/documents/mail/Illegal%20Content%20Violation%20-%20Message%2097607.txt
> X-Bogosity: No, tests=bogofilter, spamicity=0.926449, version=0.16.0
> http://files.persson.tm/documents/mail/Illegal%20Content%20Violation%20-%20Message%204703.txt
> X-Bogosity: No, tests=bogofilter, spamicity=0.000293, version=0.16.0
> http://files.persson.tm/documents/mail/Extreme%20Content%20Violation%20-%20Message%2073119.txt
> X-Bogosity: No, tests=bogofilter, spamicity=0.863759, version=0.16.0
> 
> I score the mail with the following command through a .qmail file
> /usr/bin/bogofilter -H -p -C -d 
> /webmail/8/4/84deda69dc38c4a39443e9b7a8230018/bogofilter/

Why do you use -H and -C?

Anyhow, your second message is broken. It looks like saved
from a bad reader. No surprise it does not work out. The
others get some value, so it seems to work. It is not clear
why you mean you have 1000 messages of this kind, they look
totally different.

My suggestion: Move to an up-to-date version and do a new
trianing from scratch. Using header tagging is useful, use
it. Use a training which does not require tuning or do tune
your parameters.

pi