Topic: Logterse plugin for LearnAsSpam Bayesian filtering ?

[zeeclor]

For the record.

I have recently implemented the LearnAsSpam Bayesian filtering as per the the Sonora Guide for SME "(http://www.sonoracomm.com/index.php?option=com_content&task=view&id=49&Itemid=32)

Today I get stats on the filtering process. They are all blank however with a messages saying that I do not have the logterse plugin (http://www.mail-archive.com/qpsmtpd@perl.org/msg06410.html). This is not surprising since it is a new plugin.

There is a line that says :-
"to enable it follow the instructions at ............................. "
which has a certain mystery about it, but does not help.

The logterse authors says:-
"How to use it:

- invoke logterse from the config/plugins file the same as for any plugin.

- turn your log levels right down and enjoy the silence  
- if you're using multilog do something like this:

    multilog t !/path/to/qplogsumm ./main

When the log is rotated qplogsumm will be run and its cumulative statistics will be stored in ./main/state. The stored log will remain unchanged."

but this is beyond my non-existent Perl, Qmail knowledge.

David

[compdoc]

Use the logterse and other files located here, as they are specific to SME:

http://mirror.contribs.org/smeserver/contribs/bread/mailstats/

The instructions are there as well.

Once logterse is installed, as new mail comes in, you should see lines added to the /var/log/qpsmtpd/current  file that say 'logterse plugin:' followed by info about the email. (scroll to the bottom of the file to see new entries)

Theres no need to run the multilog stuff.

By the way, there are two forms of Bayes - one you have to teach, and one that learns spam automatically. LearnAsSpam.pl needs users to put all their spam into a folder named LearnAsSpam that you create for each user in /home/e-smith/files/users/<username> .

Then they have to connect to SME and manage it using IMAP, (if they use a client like Outlook) or using a browser and SME webmail/Horde.

IMAP is a bit more trouble than pop3. You'll have to set it up for users, show them how to use it and where to place the spam. I use both types of Bayes, but for customers who arent as likely to want to mess with their email that much, I use the automatic form of Bayes, rather than LearnAsSpam.pl

[zeeclor]

Thanks Compdoc, that's got me up and running.