[Mailman-Users] Pipelining with Mailman?

Mark Sapiro msapiro at value.net
Tue Dec 12 01:44:21 CET 2006


matt at rocafella.org wrote:
>
>I have a question regarding pipelining, etc.  If my mailing list has 600
>subscribers from domain.com, will Mailman combined with my MTA (Postfix)
>pipeline all of these messages that are destined to the same domain?


Maybe.


>From
>what I understand,  it will do this and send all of these messages with
>one unique Message ID, instead of 300 separate Message Id's.  This should
>reduce the likelihood that the remote mail server will filter my emails as
>spam.  Does this sound right?


First, all the messages sent by Mailman for one post will have the same
Message-Id: which will be the Message-Id: from the incoming post.

The messages will be sent from Mailman to Postfix in a number of SMTP
transactions, each of which will have one MAIL FROM command and either
one or many RCPT TO commands. The number of recipients per transaction
will be one if the list is personalized or if the delivery is VERP'd.
Otherwise, the number will be as many as possible up to SMTP_MAX_RCPTS
(default = 500).

The chunks of size SMTP_MAX_RCPTS are created by grouping the
recipients (which are in a hashed order to start with) by top level
domain. All .com addresses are in one group, .net and .org are in a
second group, .edu, .us and .ca are in a third group and all the rest
are in a fourth group. The chunks are created by pulling up to
SMTP_MAX_RCPTS addresses from a group until it's empty.

Thus, depending on list size and SMTP_MAX_RCPTS, a chunk may or may not
contain all or most of the addresses in a given domain - top level
domains are grouped, but fully qualified domains are not grouped
within the top level.

Once a particular chunk is delivered to the MTA (Postfix), the MTA will
normally group all the recipients of that chunk by receiving domain
and send one transaction to the receiving domain containing all the
recipients from that chunk in that domain. The MTA will not group
recipients from the same domain from two different chunks, as it
doesn't know that the message content of two chunks is the same.


>I am basing my argument that we need to
>switch to Mailman on the fact that the old script takes ages (several
>hours) and that it is more likely to get us blacklisted because of the
>method of sending mail that it uses.


Mailman/Postfix should certainly be able to improve significantly on
the 'several hours'. Whether the receiving domain's getting one or a
few messages with several recipients each vs. a bunch of individual
messages is more or less likely to trigger the receiving domain's
blacklisting is something that probably depends on the receiving
domain and something that the receiving domain might consider a secret.

-- 
Mark Sapiro <msapiro at value.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