Okay folks, I am finally releasing Mailman 2.0 beta 3. I think I have
nailed the duplicates problem -- at least I have seen no duplicates in
the last 24 hours on python.org, so I'm feeling good about it (we'll
see if that lasts tomorrow morning :).
Please give this version a thorough test; I'd like to have a much
shorter release cycle for the next few betas. I am aiming to release
2.0 final on July 14th, the Friday before the O'Reilly OSSCON. I plan
on spending time between now and then on slogging through the
discussion lists and bug reports, and working on the documentation.
See below for an excerpt from the NEWS file for changes since
2.0beta2. If you have not been following the CVS updates and are
upgrading from 2.0 beta 2 or earlier PLEASE READ THE UPGRADING FILE!
A crucial step is to reload your crontab.in file, otherwise you won't
clear your message queue often enough (or perhaps not at all).
-------------------- snip snip --------------------
2.0 beta 3 (26-Jun-2000)
- Delivery mechanism (qrunner) refined to support immediate
queuing, queuing directly from MTA, and queuing on any error
along the delivery pipeline. This means 1) that huge lists
can't time out the MTA's program delivery channel; 2) it is much
harder to completely lose messages; 3) eventually, qrunner will
be elaborated to meter delivery to the MTA so as not to swamp
it. The tradeoff is in more disk I/O since every message coming
into the system (and most that are generated by the system) live
on disk for some part of their journey through Mailman.
For now, see the Default.py variables QRUNNER_PROCESS_LIFETIME
and QRUNNER_MAX_MESSAGES for primitive resource management.
The API to the pipeline handler modules has changed. See
Mailman/Handlers/HandlerAPI.py for details.
- Revamped admindb web page: held messages are split into headers
and bodies so they are easier to vette; admins can now also
preserve a held message (for spam evidence gathering) or forward
the message to a specified email address; disposition of held
messages can be deferred; held messages have a more context
meaningful default rejection message.
- Change to the semantics for `acceptable_aliases' list
configuration variable, based on suggestions by Harald Meland.
- New mm_cfg.py variables NNTP_USERNAME and NNTP_PASSWORD can be
set on a site-wide basis if connection to your nntpd requires
- The list attribute `num_spawns' has been removed. The mm_cfg.py
variables MAX_SPAWNS, and DEFAULT_NUM_SPAWNS removed too.
- LIST_LOCK_LIFETIME cranked to 5 hours and LIST_LOCK_TIMEOUT
shortened to 10 seconds. QRUNNER_LOCK_LIFETIME cranked up to 10
hours. This should decrease the changes for bogus and harmful
- Resent-to: is now one of the headers checked for explicit
- Tons more bounce formats are recognized. The API to the bounce
modules has changed.
- A written LockFile module which should fix most (hopefully all)
bugs in the locking machinery. Many improvements suggested by
Thomas Wouters and Harald Meland.
- Experimental support (disabled by default) for delivering SMTP
chunks to the MTA via multiple threads. Your Python executable
must have been compiled with thread support enabled, and you
must set MAX_DELIVERY_THREADS in mm_cfg.py. Note that this may
not improve your overall system performance.
- Some changes and additions to scripts: bin/find_member now
supports a -w/--owner flag to match regexps against mailing list
owners; bin/find_member now supports multiple regexps;
cron/gate_news command line option changes; new script
bin/dumbdb for debugging purposes; bin/clone_member can now also
remove the old address and change change the list owner
- The News/Mail gateway admin page has a button that lets you do
an explicit catchup of the newsgroup.
- The CVS repository has been moved out to SourceForge. For more
information, see the project summary at
- Lots 'o bug fixes and some performance improvements.