[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: WAS: How to feed Bayes on relay-only server?

From: Dan Nelson
Subject: Re: WAS: How to feed Bayes on relay-only server?
Date: Mon, 14 Jun 2004 23:04:15 -0500
User-agent: Mutt/1.5.6i

In the last episode (Jun 14), address@hidden said:
>  You said:
> > .. save all incoming messages under 32k 
> > .. agent that grabs just the Message-ID
> > .. submits them via xmlrpc to a daemon on .. server
> > .. extracts the full messages from the database
>  Have you considered releasing this stuff to the spamassassin
> people/users ... any docs on what you are doing or have created?
>  I have linux based S-A servers acting as frontends to exchange
> servers at a few locations. What you are doing sounds very good. I am
> not a programmer but i can follow "recipes" rather well! :-) Do have
> about 10 years linux experience though.
> -Patrick Main-

It's not quite recipe-level, unfortunately.  The whole user-interface
part would have to be written for Outlook, for example, and you may
want to rewrite the pike script in a language you're more familiar with
(perl, php, vb).  I picked xmlrpc as the protocol because it's
relatively portable across systems and languages, and supports complex
data types like arrays and structs.  Here's the 5-minute tour.

The message-saving milter is in CVS, under the "save-milter" directory. 
It currently expects a mysql daemon at a host called "spamd", and a
"mail.messages" table with a layout like the "schema" attachment. Every
hour I run the "cleanmail" attachment to age the database.  

There's no easy way to export a Lotus Notes agent and GUI to text, but
that won't help you much with Outlook.  You're on your own there..  The
xmlrpc client is in the "rpc.cgi" attachment, written in a C-like
scripting language called Pike.  Most of it is fluff to handle training
both of the clustered servers and format the output text for the
end-user.  I can't wait for spamassassin to go all-SQL for its bayes
storage :)  The cgi expects to be passed an xmlrpc function called
learnspam(mids, flags).  mids is an array of message-ids you want to
train, and flags is a struct where you pass the end-users name,
ham/spam flag, etc.

        Dan Nelson

Attachment: schema
Description: Text document

Attachment: cleanmail
Description: Text document

Attachment: rpc.cgi
Description: Text document

reply via email to

[Prev in Thread] Current Thread [Next in Thread]