[Mailman-Users] The moderation/SPAM queue

Barry A. Warsaw barry at zope.com
Fri May 17 23:35:22 CEST 2002


>>>>> "AcLA" == AerosmithFanClub com List Admin
>>>>> <listadmin at aerosmithfanclub.com> writes:

    AcLA> 	Now thats about the stupidest statement I have ever
    AcLA> seen!  Mail that you dont see on the list does not get to
    AcLA> the list.  Same damn thing.

Dude, please chill.  I think we can get a lot more accomplished on
this list if we respect each other.
    
    AcLA> What people send is another issue.  Folks can send what ever
    AcLA> they want where ever and when ever they want.  As long as it
    AcLA> is blocked from the list then the list does not get the
    AcLA> mail.  I stand by what I said.  If you set up your list
    AcLA> correctly then the list will not get any spam which means
    AcLA> none of the list subscribers will get any spam in their
    AcLA> mailboxes that are a result of being subscribed to the list.

I firmly believe that spam fighting is not Mailman's primary function
in life.  It has some minimal blocks, but I see those as being mostly
ineffective.  That doesn't bother me much though since I believe there
are really excellent spam blocking add-ons, most notably SpamAssassin.
Mailman will keep its built-in blocks because it needs to provide
something out of the box.  But in the same way that 3rd party archives
can be slotted in for better archiving support, I'd like to see a
similar architecture for spam blocks.

There's a number of ways to set up effective spam filters.  You can
set up something like SA at the MTA level, filtering email before it
ever gets to Mailman.  That's what we do on python.org and it's great
because it blocks spam not just to the lists, but to every address
@python.org.

Or you can integrate SA with Mailman so that the spam determination is
part of Mailman's normal moderate-and-munge processing.

Personally, I think the most promising approach is a combination of
the two.  Have SA filter between the MTA and Mailman, automatically
dropping anything with score > N.  For M < score <= N, pass them on
and have Mailman hold them for approval.  For scores 0 < score <= M,
they should get passed through unmodded.

Finding the right values for M and N may be tricky, but from our
experience I think something like 7 or 8 for N is reasonable and
something like 4.5 or 5 for M is reasonable.

-Barry






More information about the Mailman-Users mailing list