It is pretty clear here that we are talking about a configuration that is
actually specfically prohibited by 2821.
If you are doing SMTP and claiming to be 2821 compliant you must lookup the MX
and you must not look at the A if there is no MX there. Any sender that is
breaking those rules is not compliant with the spec.
Spam is relevant in this regard to the extent that when you have a protocol
that is under such a level of sustained attack, it is entirely justified for
receivers to employ strict compliance with the standard as an acceptance
criteria. If you are not 2821 compliant then no soup for you.
I see no reason at all to extend support for non-compliant systems to the IPv6
world. On the contrary, it seems to me that this transition is exactly the
point at which you would want to say that the host name fallback support is
The argument would become even stronger if it turned out that email comming
over IPv6 transports could be more easily distinguished from spam than over
other transports. This might even be turned into an incentive for making the
What I am thinking here is that we stand a better chance of authenticating IPv6
address block allocations and thus eliminating bogons. Residential support for
IPv6 is non-existent today. We have a good chance of getting some BCP type
agreements out of the likes of MAAWG that could limit botnet and spambot
Lets make IPv6 as clean as possible, leave host name A record fallback to the
From: ietf-bounces(_at_)ietf(_dot_)org on behalf of Ned Freed
Sent: Wed 26/03/2008 2:48 AM
To: Keith Moore
Cc: John C Klensin; Frank Ellermann; Ned Freed; ietf(_at_)ietf(_dot_)org; Bill
Subject: Re: Last Call: draft-klensin-rfc2821bis
It might be the case that it's useful for an MTA to have an option to
skip MX lookup for specific destinations because of DNS brokenness at
those destinations. But this seems to me to be outside of the scope of
By the same token, discussions of gatewaying to non-Internet systems could be
considered "outside the standard". But RFC 2821 devotes many pages to
discussing this sort of thing.
Skipping MX lookup is not acceptable as a general
practice, nor is it something we want to encourage.
I never implied that it was acceptable. In fact I'm fairly sure I said
the exact opposite.
In general, it's always been acceptable to configure an MTA to handle
mail in some special-case way for specific domains where there was
specific knowledge such that the special-case handling made sense for
those domains. The MX-then-A lookup is what you should do in the
absence of any such knowledge.
IETF mailing list
IETF mailing list