[Mailman-Users] GNU Mailmain 2.1.9: Filtering Messages based on

Hank van Cleef vancleef at lostwells.net
Wed May 21 19:11:03 CEST 2008

The esteemed Steve has said:
> I'm admiring a GNU Mailman email list that is running on version 2.1.9
> on a server that I do not have access too.
> Is it possible through the use of the admin interface to filter out
> messages based on key words in the body of an email?
> I have some banned ex-subscribers who periodically resubscribe under
> new email addresses to post URLs to rant pages about my list.    I
> would rather not moderate all new incoming subscribers.   I don't want
> to slow legitimate users down.  The trolls often bury the URLs at the
> end of long messages making it easy to miss.
I'll comment on this request from an administrator's perspective.  If
there is a software solution to the problem of rogues subscribing to a
list and using it to disrupt things, I do not know what that would be.
Rogues are very good at finding a software trap pattern and changing
their tactics to bypass it.  

Historically, the main list I administer was set up on a listserv
system with subscription "by invitation only" in 1998.  It was
purposely set up to get away from another list that was dominated by
trolls, spammers, and flamers.  We migrated the list to another
system, using Mailman, in 2001.  Over the years we have endured
several concerted attempts by those who made themselves unwelcome and
who were unsubscribed, to resubscribe under false pretenses and do a
hit-and-run attack on our list.  For several years, there was another
mail list, hosted elsewhere, populated by "our enemies," but I think
that has more-or-less fallen by the wayside.   

What has worked for us is a two-fold administrative solution.  
1.  We do not do software subscription through the Mailman interface.
All subscriptions are manually-processed.  New subscribers are
required to fill out a form that requests real name, address,
telephone number, educational level, profession, and birth year and
mailed to the moderators through a cgi-bin script.  The moderators
review these and manually subscribe those who "pass" through the bulk
subscription Mailman resource.

2.  All new subscriptions are moderated.  Once we see a few clean
posts from new subscribers, we take them off moderation.  Virtually
all of the rogues who have managed to get through the subscription
screening tip their hand quite quickly.  Some offlist e-mail between a
moderator and an offender either gets an apology or a flame.   

Which leads to:

3. Mail lists post messages from humans that are intended to be read
by other humans.  What is acceptable behavior is strictly human in
nature.  Sociology, rather than technology, is what will set and
maintain standards.  It is very important to have moderators who have
good skills at guiding group behavior.  And, yes, there is moderator
workload involved in making this work.  Once a list's focal point and
demographics are clearly established, almost all issues involving
listmember conduct are sociological.

> I would prefer not to use procmail or software other than GNU Mailman.
>  As I wrote the list is made available to me freely on a server that I
> do not have access to.  I don't want to bother my admin more than I
> have to.
Our use of various filter programs is oriented toward reducing spam
volume.  We do not attempt to use it to detect rogue posts.

> Is it possible to set up GNU Mailman 2.1.9 to filter on URLs or other
> keywords in the body of an email?
What it is possible to do is one thing.  What is practical and
effective is another.  I've worked in the field of sociology of
computer-mediated communications for about thirty years now, and have
coded up no end of "good ideas" that more-or-less failed in their
objective of reducing human workload.  

Mailman is rich in resources to handle basics, such as reformatting
funny-format MUA text into readable text, stripping off attachments
(with or without virii), not allowing postings from non-registered
people, and the like.  But it does not attempt to use technology to
regulate human behavior.  My experience suggests that if one is going
to run a "clean" mail list, one has to define "clean"---in human
sociological terms.  And it is going to take humans to set and enforce
those standards.  Ten years of experience with our list has also made
clear that "clean" is a moving target.  


More information about the Mailman-Users mailing list