[Mailman-Users] Bounces Issue - Revisit II

Barry A. Warsaw bwarsaw at cnri.reston.va.us
Fri Dec 17 05:53:28 CET 1999


Mailman 1.2 (or will it be called 2.0? :) will improve bounce
processing considerably.  I've rewritten the bounce detector as a
pipeline, similar to the new message delivery pipeline.  I've also
added new bounce detectors for quite a number of common formats
(qmail, postfix, dsn, yahoo, ...).  I've got things set up on
python.org so that I see every missed bounce that my list admins see,
so at the very least, I'll be adding them to the test suite.  Trust
me, this incentivizes me to nail the most common bounce formats :)

There's ultimately no way to catch 100% of the bounces.  There are
just too many brain dead MTAs out there. :(

>>>>> "DB" == Dan Busarow <dan at dpcsys.com> writes:

    DB> It's an announce only list, the owner fired off a message and
    DB> 5 minutes later the machine was running out of swap because of
    DB> all the python processes handling bounces.  Only about 1500
    DB> messages had gone out and we had in excess of 100 python's
    DB> running. (P400 w/ only 64M RAM).

In Mailman 1.2, you'll get one Python process per bounce you receive.
No way around that, but at least it doesn't fork like crazy any more.
JC addressed this issue well in his response.

If you feel adventurous you can always try to the current CVS
snapshot.

-Barry




More information about the Mailman-Users mailing list