Great Circle Associates List-Managers
(May 1997)

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: rejected mail - RFC822 conflict ???
From: Brad Knowles <BKnowles @ aol . net>
Organization: America Online, Inc.
Date: Wed, 07 May 1997 13:51:36 -0400
Cc: Valdis Kletnieks <Valdis . Kletnieks @ VT . EDU>, Michael Ramundo <sysmrr @ CNSIBM . ALBANY . EDU>, Jeff Kell <jeff-kell @ UTC . EDU>, Pete Weiss <Pete-Weiss @ PSU . EDU>, list-managers @ GreatCircle . COM, Eric Thomas <ERIC @ VM . SE . LSOFT . COM>
Reply-to: KnowlesB @ aol . net


    I've been informed that there is some discussion going on in
these two lists regarding our refusal to accept source-routed mail.

    Well, it's true.  We do refuse to accept email with source-routed
envelope addresses.

    I also know what RFC 1123, section 5.2.19 says.  I'm also
on the IETF DRUMS WG, and I know what the upcoming drafts
say with regards to refusing mail (see section 7.5 of
I quote:

	It is a well-established principle that an SMTP server
	may refuse to accept mail for any operational or technical
	reason that makes sense to the site providing the server.

    Specifically, source-routed mail has historically been the
source of no end of problems, and frequently abused by less savoury
types in attempts to ensure that they don't have to deal with their
bounces, etc..., we consider this an operational issue and will
refuse to accept mail with source-routed envelope addresses.

    Our general approach is one whereby if we can't determine that
we could bounce a message if we had to (i.e., a message comes in
for a nonexistant AOL user, or for one whose mailbox is full), then
we won't accept that message.

    The standards are much more forceful on the issue of accepting
full responsibility for a mail message once you've accepted the
message itself, and that responsibility extends to delivering error
messages back to the sender if there is some sort of problem.

    At some point in time, the RFC 1123 "Robustness Principle"
(section 1.1.2) continues the propagation of more and more bad
systems, because properly behaved systems are required to continue
to be tolerant of poor behaviour, thus ensuring that we'll *never*
free ourselves from those shackles.

    Essentially as much has been observed by various Internet mail
experts, many of whom are working on drafting the upcoming standards.
I know that I will personally work to ensure that source-routed
email is deprecated as much as possible, preferably to the status of
"MUST NOT generate/MUST NOT accept".

    In the meanwhile, AOL will refuse to accept email with
source-routed envelope addresses, and I'll contribute my
rewrite rules to Eric Allman in hopes that I can convince him to
incorporate them in an upcoming release of sendmail (at least as
a FEATURE() you can turn on, if not turned on by default).

    As Jeff Kell <jeff-kell @
 UTC .
 EDU> noted on LSTOWN-L, since
you can change the configuration of ListServ so that it doesn't
generate source-routed email messages, this shouldn't pose too
big of a problem for you.

    I'd like to see this default changed in upcoming releases of
ListServ, so that in the future, email with source-routed envelope
addresses will not be generated unless you explicitly configure it
to do so.

Brad Knowles                                MIME/PGP: KnowlesB @
 aol .
    Senior Unix Administrator              <>

Brad Knowles                                MIME/PGP: KnowlesB @
 aol .
    Senior Unix Administrator              <>

Indexed By Date Previous: help
From: "Ljubisa Gavrilovic (LJ)" <lj @ dream . vol . net . mt>
Next: Threats from an unsubscribed list member
From: Graeme Read & Grant Whittingham <whitread @ ozemail . com . au>
Indexed By Thread Previous: help
From: "Ljubisa Gavrilovic (LJ)" <lj @ dream . vol . net . mt>
Next: Re: rejected mail - RFC822 conflict ???
From: Keith Moore <moore @ cs . utk . edu>

Search Internet Search