[Mailman-Users] Mail delivery gets congested

Mark Sapiro mark at msapiro.net
Thu Nov 10 14:59:04 EST 2016


On 11/10/2016 10:49 AM, Christian F Buser via Mailman-Users wrote:
> 
> On September 2, I asked here about a problem that subsciber's email addresses were displayed in the message headers, and you advised me to change the delivery method of Mailman to "Full Personalization". 

What I said on Sept 2 was:

> change Mailman to send only one recipient per transaction. This can
> be accomplished in multiple ways:
>   a) If personalize is available in Non-Digest options, set it to Yes.
>   b) add
> VERP_DELIVERY_INTERVAL = 1
>      to mm_cfg.py
>   c) add
> SMTP_MAX_RCPTS = 1
>      to mm_cfg.py
> 
> (I recommend b)


See
<https://mail.python.org/pipermail/mailman-users/2016-September/081248.html>.

Note that personalize = Yes is not the same as personalize = Full
Personalization, although they will both solve your address exposure
problem and both have the same effect  on SMTP transactions.

The difference is Full Personalization also changes the To: header of
each delivered message to the recipient's address whereas Yes leaves it
unchanged.


> Having one this, it seems that Mailman is now having problems to send out the messages within a useful time frame. The porvider told us that SMTP is currently set to accept only 30 messages per individual connection. So with 200 subscribers it would take 7 connections to get out one single message.  I do not know about the time between two connections, but I saw that it takes at least 20 minutes and in several cases has taken more than one out until the messages are out. 


Actually the whole point of any of the changes I suggested was to cause
Mailman to send only one recipient per SMTP transaction because your
original issue was apparently due to the MTA adding an Envelope-To:
header with all the transaction's recipients.


> The provider said that he can increase the numnber of messages per connection but thinks that this is also a security issue (spam) and the amount can only be set server-wide (no possibility to make an exception rule for mailing lists). 


That's irrelevant at this point because you are sending to only one
recipient per transaction.

The issue is the MTA is taking too much time to process a transaction
and your SMTPD delivery for 200 recipients has gone from about 7
transactions with about 30 recipients each to 200 transactions with 1
recipient each. You needed to do this because of the MTA's putting all
the transaction's recipients into the Envelope-To: header.


> What could I do from my side to increase delivery speed? 


Set personalize = No to go back to the original issue.

This has to be addressed in the MTA configuration. Only the host can do
that.

There are many cPanel Mailman installations with much bigger lists that
200 members that don't have these issues.

If your host is willing, they can join this list and ask for help here
and we will do our best to help. They may also benefit from
<http://www.exim.org/howto/mailman21.html>, although there are cPanel
specific differences in the Mailman router and transport which are
required for cPanel's virtual domain support. The section at
<http://www.exim.org/howto/mailman21.html#batune> could be useful, and
of course as I suggested in my Sept 2 reply, they can remove the
envelope_to_add option from the mailman transport and quit exposing all
the recipients in an Envelope-To: header (see
<http://www.exim.org/exim-html-current/doc/html/spec_html/ch-message_processing.html#SECID225>
and the envelope_to_add setting at
<http://www.exim.org/exim-html-current/doc/html/spec_html/ch-generic_options_for_transports.html>).

As a list admin with no ability to change the host, you have only the
options of leaving things as they are and tolerating the delivery delay
or going back to personalize = No and tolerating the address exposure.

-- 
Mark Sapiro <mark at msapiro.net>        The highway is for gamblers,
San Francisco Bay Area, California    better use your sense - B. Dylan


More information about the Mailman-Users mailing list