Re: Gnus and Outlook and spam?

[Ted Zlatanov <tzz@lifelogs.com>]

On Tue, 28 Oct 2003, colman@ppllc.com wrote:

>>>>>> "TZ" == Ted Zlatanov <tzz@lifelogs.com> writes:
> 
>    TZ> Please point that out.  You may be referring to the note for
>    TZ> IMAP users, which relates to statistical filters, but that's
>    TZ> not a problem with IMAP per se, it's just a setting
>    TZ> (nnimap-split-download-body) that you need to turn on.
> 
> This variable, nnimap-split-download-body, is NOT required if I work
> with bogofilter?  Supposedly, this variable is expensive to use
> since it slows things down considerably.

Bogofilter needs to analyze the message body, so yes, you need to
download the body.  In fact, unless you specifically turn it off, any
statistical filter, e.g. spam-use-bogofilter, will turn body
downloading on.  You could, however, turn body downloading off if you
just want to analyze the message headers.  It's a user decision, in
the end.

>    TZ> ;; this is a very rough approximation, read the fancy
>    TZ> ;; splitting docs if you want to know more
>    TZ> (setq nnimap-split-fancy
>    TZ>       '(|
>    TZ>          (any "help-gnu-emacs" "INBOX/Emacs/Help")
>    TZ>          (any "info-gnus-english" "INBOX/Emacs/Gnus")
>    TZ>          ...
>    TZ>          ...
>    TZ>          ...
>    TZ>          ...
>    TZ>          (: spam-split)
>    TZ>          ("INBOX/Misc")))
> 
>    TZ> (setq spam-use-bogofilter t)
>    TZ> (spam-initialize)
> 
> Thanks for this conversion.  If I understand this correctly, it will
> first filter that which it knows, then it will spam-split and file
> accordingly.  This means that I will only be spam filtering on those
> emails that I have not explicitly filtered and filed, correct?  

Yes.

> So it cannot check for spam in mailing lists that have been spammed,
> right?

You can.  Let's say you have tests which are mild, mid-range, and
severe (examples of each are spam-use-blacklist, spam-use-bogofilter,
and spam-use-BBDB-exclusive respectively).  You can do:

(: spam-split 'spam-use-blacklist) ; mild test, will catch known spammers
(MAILING LIST CHECKS) ; put mailing lists with little spam here
(: spam-split 'spam-use-bogofilter) ; mid-range test, will catch suspected spam
(OTHER CHECKS) ; you can put mailing lists that get spam here
(: spam-split 'spam-use-BBDB-exclusive) ; severe test, everyone not in the BBDB is a spammer
"mail"

The exact structure is up to you.  Most people don't need more than 3
checks.  I personally do BBDB, blackholes, and regex (which catches
server-side SpamAssassin tagging).  My nnimap-split-fancy:

'(|
 (: gnus-registry-split-fancy-with-parent)
 (: spam-split 'spam-use-regex-headers)
 ;; lists are filtered here, all but one omitted for brevity
 (any "ding" "ding")
 ;; this will redo the regex-headers check, but that's a cheap test
 ;; so I don't mind
 (: spam-split) 
 "mail")

Ted