On Fri, 2002-10-11 at 07:20, Federico Sevilla III wrote:
> The available documentation only seems to be useful for systems wanting
> their users to have individual bogofilter wordlists. I also read Paul
> Graham's "A Plan for Spam", and he recommends this kind of setup, as
> well.
i noticed something when i tried to add non-spam words. the man
page says to use -n (and -N) to add non-spam words. this is wrong.
in the source it's actually -h and -H. trivial to edit main to
handle -n and -N though.
the main problem with using a global wordlist is, not
everyone agrees on what is spam and some people might actually
*want* to receive all those ads for pr0n. also, except for
tweaking individual user's .procmailrc (and i don't even
like tweaking my own, fire and forget is what i want), there
doesn't seem to be any good way to do whitelisting for
bogofilter so that email that is tagged but is actually
"good" isn't flagged as spam.
with respect to auto-deleting or quarantining spam. that's
another hard problem. it's possible for false positives to
get lost if we auto-delete or quarantine on the server (and
then give the user complicated [because of security, so that
only the right users get to view their own spam] procedures
for looking at their spam so they can decide if they really
want it.
bogofilter has one advantage to systems like razor2 though
(which are medyo sensitive to the full content of the email
and not just to the words in it) you can safely forward
spam to a bogofilter email address (or to a human who will
review the spam before adding it to the bogofilter wordlist).
it doesn't get confused by the effects of forwarding (>,>>,
--Forwarded Mail--, etc). so if users want to, they can
report spam to be added to the wordlist just by forwarding
to an email address.
> I'm curious, anyway: would it be too much to ask (plead with?) you to
> detail how you hacked bogofilter to act on system-wide wordlists and
> hooked it into the Postfix pipeline? What other components do you have
> in this pipline? How is mail "quarantined"?
i am not orly :). no doubt he'll tell you how he did his. and
in fact, i haven't even implemented bogofilter systemwide yet since
i'm still testing (and it's not yet working well enough, my wordlists
are puny compared to orly's, only 25MB of spam) but see above
for systemwide wordlist (or you could edit the source, see BOGODIR
(approx line 21) in main.c and in the main function, he sets up the
directory variable using getenv("HOME") and BOGODIR (approx line 40).
just modify those to point to a global place.
with respect to the postfix pipeline, well, orly will have to do
that. i haven't gotten that far yet, and my ISP uses sendmail
(not my choice, no sendmail vs postfix debates please. that's been
done).
autodaemon looks good. i haven't used it yet. but that's a good
way to go to cut down on the expense of starting up bogofilter.
if only i could figure out a good way :). orly, did you use
autodaemon? or did you just paste some of the bogofilter code
into one or another of the postfix daemons?
tiger
--
Gerald Timothy Quimpo tiger*quimpo*org gquimpo*sni-inc.com tiger*sni*ph
Veritas liberabit vos.
_
Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph
To leave: send "unsubscribe" in the body to [EMAIL PROTECTED]
Fully Searchable Archives With Friendly Web Interface at http://marc.free.net.ph
To subscribe to the Linux Newbies' List: send "subscribe" in the body to
[EMAIL PROTECTED]
