New Bayes DB install

[Joe Borg <jo...@datastream.com.mt>]

Hi,
I'm still pretty much a newbie to Spamassassin so please excuse me if I'm
asking an already answered questions; however, I've searched the site for an
answer and could not find one.

I'm currently running Spamassassin v. 2.63 and would like to upgrade to the
latest version in the coming weeks. I've read the upgrade instructions,
however, I would not like to retain my old bayesdb since I fear that I
incorrectly trained it. Is it possible to upgrade but force spamassassin to
install a new/fresh bayes db?

Thanks,

Joe

Re: New Bayes DB install

[Loren Wilton <lw...@earthlink.net>]

> >From spamrep@mydomain.com  Tue May 17 09:54:07 2005
> From: "winfred fuller" <da...@hates.every1.net>
>
> Where:
> spamrep: is my Pc-Pine email account
>
> The only thing that worries me is the very first 'From' line since this
must
> have been added in bouncing. Can I tell bayes to ignore this line in
anyway
> without it also ignoring the 'From:' line (note the difference).

Can't tell Bayes to ignore anything as far as I know, beyond the normal SA
headers which will be stripped out.  So you will have to find a way to
delete this line as part of stripping out the attachment, I suspect.

It may be that if that is a constant header and doesn't ever show up in
normal mail it won't end up biasing Bayes anyway (it will just learn a few
tokens it will never see in real mail).  So you may not have to deal with it
anyway.

        Loren

RE: New Bayes DB install

[Joe Borg <jo...@datastream.com.mt>]

The general rule (I'm inclined to say the absolute and only way) with
Outlook and OE is to set up a folder, typically IMAP, and share it as a
public folder to the clients,  They can then drag&drop, or
rightclick-and-Copy/Move the message into the ham or spam folder.

You then harvest the IMAP folder(s) with some cron script and feed them to
SA, or possibly by hand if you want to scan the stuff and make sure the
users have a clue about what is ham and what is spam.

Anything that requires forwarding or similar will screw up the message
beyond usability.  In theory forwarding *as an attachment* and then
stripping the attachment out *should* work - but a number of people have
said that Outlook (but not OE) screws this up too.

A number of people have posted scripts or links to scripts to automate the
learning process with this sort of a setup.

        Loren


Thanks for the great info. One final question. In view of the outlook
problems, I've recently installed PC-Pine on my PC and instructed all users
to forward undetected spam as an attachment to my PC Pine email. I then
bounce the actual attachment to SA, taking care of the ReSent headers in the
local.cf file. Should there be any problems with this setup? Mny only
concern is that in the learned-spam file on my Sa server, headers show up as
follows:

>From spamrep@mydomain.com  Tue May 17 09:54:07 2005
Return-Path: <sp...@mydomain.com>
Received: from localhost ([217.15.97.57])
        by mailserver.mydomain.com (8.12.11/8.12.11) with ESMTP id
j4H7s5oC007677
        for <sp...@mydomain.com>; Tue, 17 May 2005 09:54:07 +0200
From: "winfred fuller" <da...@hates.every1.net>

Where:
spamrep: is my Pc-Pine email account
spamtrap: is the account to which messages are bounced.

The only thing that worries me is the very first 'From' line since this must
have been added in bouncing. Can I tell bayes to ignore this line in anyway
without it also ignoring the 'From:' line (note the difference).

Thanks,

Joe

Re: New Bayes DB install

[Loren Wilton <lw...@earthlink.net>]

> Incidentally, would you have any recommended way of training SA once I
> install the latest version?
>
> My install is a site-wide install with users using pop3 to check their
mail
> via windows clients (typically outlook). My concern here is that in
bouncing
> messages to SA; I do not want the learning process to use any of the added
> headers as a shortcut. Currently I have this problem since I was using
> outlook to bounce messages and SA ended up using my own email address as a
> shortcut...

The general rule (I'm inclined to say the absolute and only way) with
Outlook and OE is to set up a folder, typically IMAP, and share it as a
public folder to the clients,  They can then drag&drop, or
rightclick-and-Copy/Move the message into the ham or spam folder.

You then harvest the IMAP folder(s) with some cron script and feed them to
SA, or possibly by hand if you want to scan the stuff and make sure the
users have a clue about what is ham and what is spam.

Anything that requires forwarding or similar will screw up the message
beyond usability.  In theory forwarding *as an attachment* and then
stripping the attachment out *should* work - but a number of people have
said that Outlook (but not OE) screws this up too.

A number of people have posted scripts or links to scripts to automate the
learning process with this sort of a setup.

        Loren

RE: New Bayes DB install

[Joe Borg <jo...@datastream.com.mt>]

-----Original Message-----
From: Duncan Hill [mailto:satalk@nacnud.force9.co.uk] 
Sent: 23 May 2005 17:07
To: users@spamassassin.apache.org
Subject: Re: New Bayes DB install

On Monday 23 May 2005 16:02, Joe Borg typed:
> incorrectly trained it. Is it possible to upgrade but force spamassassin
to
> install a new/fresh bayes db?

>> Remove the bayes files and SA will re-create them.

Thanks for the clarification :)

Incidentally, would you have any recommended way of training SA once I
install the latest version?

My install is a site-wide install with users using pop3 to check their mail
via windows clients (typically outlook). My concern here is that in bouncing
messages to SA; I do not want the learning process to use any of the added
headers as a shortcut. Currently I have this problem since I was using
outlook to bounce messages and SA ended up using my own email address as a
shortcut...

Joe

Re: New Bayes DB install

[Duncan Hill <sa...@nacnud.force9.co.uk>]

On Monday 23 May 2005 16:02, Joe Borg typed:
> incorrectly trained it. Is it possible to upgrade but force spamassassin to
> install a new/fresh bayes db?

Remove the bayes files and SA will re-create them.