On Fri, Feb 28, 2003 at 08:54:10AM -0500, Frank Pineau wrote:
> > On Fri, Feb 28, 2003 at 12:23:43PM +1100, Bill Clarke wrote:
> > > attached is a small perl script that does essentially what sa-learn
> > > --mbox does, except that the input (stdin) is a single message (like
> > > --single) which contains several message attachments. these attachments
> > > are stripped out and learnt as per sa-learn. non-message attachments
> > > are ignored.
>
> I haven't seen the parent message with the script yet, so bear with me
> please.
I received it more then 12 hours ago. Last time SA's archive at sf.net has
been updated is 3 days ago... I may resend it privately to anybody asking it
privately.
> Background: I use individual (non-global) bayes db's for my users. SA is
> called via procmail and spamd (individual procmail recipes, since not all my users
> want SA functionality). Bayes autolearn is currently working
> fine. Sendmail is my MTA.
>
> How well will this script work for users to submit their own spam/ham to
> train their own db's? I think probably each user might have to have their
> own ham/spam address pair (i.e. bob-spam@ and bob-ham@). Or would it work
> with a global address pair? How will they have to submit messages?
> Forward? Forward-as-attachment?
As shown in original post, this script is used through procmail. I use it
in my own ~/.procmailrc . For any user, let's say [EMAIL PROTECTED], any mail
to [EMAIL PROTECTED] is going to the same user.
Here is how I call SA from my ~/.procmailrc file :
:0 c:
/tmp/blurp
:0
*
^TODenis.Ducamp\+MySuperSecretPasswordAgainstSpammers-sa-learn-\/(ham|spam|forget)
| /tmp/saCVSusr/bin/sa-learn-attach --$MATCH
:0fw
| /tmp/saCVSusr/bin/spamassassin
:0:
* ^X-Spam-Level: \*\*\*\*\*\*\*\*\*\*\*\*\*\*\*
Mail/spam-almost-certainly
:0:
* ^X-Spam-Status: Yes
Mail/spam-probably
(I logrotate my /tmp/blurp everyday and keep a few days backup in case
spamassassin become crazy, but I didn't have to use it yet :)
And now I just have to "Forward-as-attachment" my spam/ham respectively to
[EMAIL PROTECTED] /
[EMAIL PROTECTED] to
add it to my Bayes db as spam/ham.
>From my ~/.maillog :
>From [EMAIL PROTECTED] Fri Feb 28 14:11:49 2003
Subject: [EMAIL PROTECTED]: 10 months of non-stop masterbation]
Folder: /tmp/saCVSusr/bin/sa-learn-attach --spam 34635
Learned from 1 messages.
>From [EMAIL PROTECTED] Fri Feb 28 14:12:49 2003
Subject: [EMAIL PROTECTED]: {** SPAM 07.10/05.00 **} Sa
Folder: /tmp/saCVSusr/bin/sa-learn-attach --spam 74278
Learned from 7 messages.
>From [EMAIL PROTECTED] Fri Feb 28 14:13:37 2003
Subject: [EMAIL PROTECTED]: {** SPAM 22.50/05.00 **} Awesome daily off
Folder: /tmp/saCVSusr/bin/sa-learn-attach --spam 103617
Learned from 11 messages.
> My goal is to eliminate the need for me to copy and paste the message from
> my desktop (my mailbox, when they forward the message to me) to my ssh
> terminal, su to the user and train their db by hand.
If your users can Forward-as-attachment (as a message/rfc822 attachment)
then they can do it themselves, they just have to use a real MUA...
Denis Ducamp.
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
