[aprssig] Mailman considered harmful - list bounce settings are bad

Matti Aarnio oh2mqk at sral.fi
Tue Jul 7 06:03:59 EDT 2009


This list, like so many others around the world, are using
a heap of Python garbage called Mailman, whose designers 
may have some clue of how email lists must work, but its users
( = list admins ) generally speaking have very little, and
therefore they leave bad default settings on as is.

One of the more harmful ways of how Mailman works is its "helpful"
auto-removal of subscribers when it decides that "there has been
excessive number of bounces".

Yeah, great..   These days that happens to large subscriber
populations, when list receives spams.
("But this is subscribers only list, it can not happen!"
Sure, standard spammer practice is to fake source addresses,
some even smartly pick correspondent pairs - like X has
posted on address Y.)

Other classes of junk are produced by programs that can not
be ordered to send plain text only, and force-feed that pesky
HTML, which is very rare in legitimate emails, and extremely
common in all manners of spams and scams.
Popular anti-junk softwares reject those too.

TAPR managers know of nothing better, but perhaps they could
be convinced to disable that thrise damned auto-removal, or
raise its trigger level to at least 10+ messages in succession.
List maintainer gets more work, but at least there are less
FALSE removals.

I know of no truly well functioning gui-controlled list
maintenance software that does not have same failings as
Mailman.  Everybody seem to have copied same bad ideas.

Reading mailman list archives, this appears to be common
topic, and relates on questionable list or system configurations:

  What are the bounce processing settings for this list. It seems
  like you may have bounce_you_are_disabled_warnings set to 0 and
  bounce_score_threshold set to <=1.0 so that members are unsubscribed
  on the first bounce. Then if some glitch occurs in SMTP delivery
  causing a bounce for all recipients, everyone is unsubscribed.

At least do raise that  bounce_score_threshold  value to 10 or more.
It does not, unfortunately, track that only a subset of messages
did bounce, it just counts received bounces.  Therefore the value
MUST be high.

Throwing away subscribers because list does occationally carry junk
is just very bad list management.  (The lists WILL carry junk,
only way to avoid that is to require posters to PGP or S/MIME
sign their posts, and accept only ones with valid signatures that
are list members...)


  73 de Matti, OH2MQK

PS: I do run largeish lists, 5300 recipients on largest one, 500+
    messages a day.  No gui anywhere, only commandline UNIX tools.
    Heaps of false rejects, tons of real rejects, human intelligence
    to separate weed from chaff.  Volumes of Aprssig are nowhere
    on that level.




More information about the aprssig mailing list