[Mailman-Users] Disabling VERP

Richard Barrett r.barrett at openinfo.demon.co.uk
Thu Feb 6 00:00:37 CET 2003


I think there is some confusion here about what is meant by VERP'ed address.

In MM 2.0.x outgoing mail from lists came from the 
<listname>-admin at mailman.domain.com alias.

With MM 2.1 outgoing mail from lists normally comes from the 
<listname>-bounces at mailman.domain.com alias.

This isn't VERP'ed mail. It is just a change in originator alias which goes 
along with some other changes aimed at improving the subsequent handling of 
any outbound email from Mailman that bounces.

 From your mail log it doesn't appear that the outgoing mail from your 
system is being VERP'ed.

With VERP'ing each outgoing mail is specific to the destination email 
address and encodes that destination in the originator email alias. So, for 
instance, mail to me from a list would be from 
<listname>-bounces+r.barrett=openinfo.demon.co.uk at mailman.domain.com. Note 
how my email address of r.barrett at opeinfo.demon.co.uk is encoded in that 
originator email alias.

The purpose of Verp'ing is to deal with the situation where mail to someone 
passes through a MTA that rewrites the destination and that subsequently 
bounces. Without VERP'ing the message sent back to mailman has lost the 
original destination. Mailman only sees the rewritten destination in the 
returned mail and cannot correlate that with any known subscriber of the 
list and hence handle the bounce appropriately. With a VERP'ed sender the 
original destination is embedded in the address to which the bounced email 
is returned. Thus regardless of re-writing of the destination of mail going 
out, if it is returned Mailman can reliably identify the subscriber whose 
mail is being bounced.

Does that resolve the problem for you?



btw: There would normally be a minimum of four entries in a sendmail log 
associated with a message posted to a list if the Mailman local sendmail 
instance is alos the SMTP server being used to handle the outgoing mail (I 
cannot speak to other MTAs as I do not use them):

1. a 'from' entry for the message being received by the Mailman-local 
sendmail instance from some other, relaying, MTA

2. a 'to' entry for sendmail delivering the message to Mailman. This will 
have the same message identifier string as 'from' entry (1)

3, a 'from' entry for sendmail sending the posting out to one or more 
subscribers. There may be several of these depending on the number of 
different domains the outgoing mail is being sent to, the number of list 
subscribers and othe factors. If VERP'ing is being applied there will be 
one of these entries for each list subscriber.

4. 'to' entries having the same message identifier string(s) as the 'from' 
entries (3) for each transfer of the outgoing mail to some other, relaying, MTA

If you have the patience and access you can track mail from source to 
destination by matching the seemingly random identifier strings. Its pretty 
much a 100% audit trail for all mail that flows. Take a look at the full 
headers of this eamil when you get it and you can see how it got from me to 

The Mailman post and smtp logs will have entires that tie in with those in 
the sendmail logs.

At 19:08 05/02/2003, Tom Maddox wrote:
><snipped massive previous discussion>
> > It looks like the code that initiates VERP'ed return addresses is in
> > $prefix/Mailman/Handlers/ToOutgoing.py and is enacted in
> > $prefix/Mailman/Handlers/SMTPDirect.py.
> >
> > I cannot see anything in those modules that would lead to
> > VERP'ing without
> > one of the appropriate variable being true.
> >
> > What exactly is being logged by your SMTP host for the messages
> > which your
> > users complain of?
>It looks like two messages get logged by sendmail for a given email:
>Feb  5 10:55:46 mail2 sm-mta[49256]: h15Itjct049253:
>ilman post listname", ctladdr=<listname at domain.com> (26/0), delay=00:00:01,
>xdelay=00:00:00, mailer=prog, pri=30914, dsn=2.0.0, stat=Sent
>Followed by:
>Feb  5 10:55:47 mail2 sm-mta[49261]: h15Itlct049261:
>from=<listname-bounces at domain.com>, size=1694, class=-30, nrcpts=10,
>9300CAE4 at domain.com>, proto=ESMTP, daemon=MTA, relay=localhost []
>In that same time-frame, the following message appears in Mailman's "post"
>Feb 05 10:55:48 2003 (41415) post to listname from listname-bounces at domain.c
>om, size=1694, success
>So, before Mailman posts the message to the list, it's already rewritten the
>header of the message to appear to be from the "-bounces" address.
>Does that shed any light on the issue?

More information about the Mailman-Users mailing list