Er?

[hughes@ah.com (Eric Hughes)]

It appears that last few fields in the returned header are responsible
for the problems.  Julf's mail also indicates why cypherpunks has had
a couple of duplicate posts recently.

The offending headers are "Return-Receipt-To" and "Errors-To".  The
"Return-Receipt-To" field is triggering a reaction in some other
mailers to bounce back acknowledgement of the mail.  Now
cypherpunks@toad.com was in the "To" list, and it appears that
acknowldegement mail was sent out to cypherpunks again.  All this time
the "Received" fields are increasing.  When there are too many of
them--the number is mailer dependent, but is typically 17-20, some
mailer along the chain bounces the message.  It sees the "Errors-To"
line and sends back the bounce to penet.  My guess is that a
significant fraction of the cypherpunks list is sending anon.penet.fi
back one message each per "Return-Receipt-To".

Not all that many mailers honor return receipts, but all mailers
bounce mail with too many Received fields.  Hence the first return
receipts sent didn't generate nearly so many errors as all the bounces
from the second time the message went out to the list.

How we solve this?  Well, let's list the mailers involved in the
particular message you sent.  The first one was the anonymous remailer
at caltech.  The message from there was directed to cypherpunks, so
that's toad.com.  From there it travelled through uunet (toad.com's
mail gateway for a large amount of traffic) to somewhere in the
gza/aktis/ov group of machines.  Somewhere in there the return receipt
was generated; note the "Return-Path: <jik@security.ov.com>" field.
This mailer generated a message back to cypherpunks (toad.com) again.
One copy of this went to a machine in uci.edu, which bounced it to
penet.

I'd say that the mailer which generated the return-receipt back to
cypherpunks (assuming that happened) is the most proximate cause.
Cypherpunks was in the To: field, not the From: field, and even though
your standard reply might go to both parties (assuming the To: field
is larger than just you), a return receipt should only be propagated
to the original sender.

toad.com is a secondary cause, since the Return-Receipt-To: field
should probably not be propagated out to a mailing list, but rather
acknowledged or discarded before mailing list expansion.  Also, since
toad.com is not running reasonable mailing list software (which we
don't have), it's not detecting duplicate messages sent back to the
list and discarding them.

Eric