[Top] [All Lists]

Re: How does SMTP IPv4 and IPv6 work together

2008-04-06 23:35:34

Robert A. Rosenberg wrote:

My error in not stating my real question but assuming the caveat by implication. To reword my query (*CAPS* are the omitted caveat):

What justification do you offer to allow an IPv6 MTA to run *VIA IPv6* without an EXPLICATE MX record? IOW: If it is dual stack and willing to use IPv4 then just allow the A-Fallback but why allow AAAA-Fallback to allow it to talk via IPv6? Note: I was assuming that an EXPLICATE MX would point not only to all the needed AAAA records but also the A records that would be used by a IPv4-Only Stack in Fallback mode (ie: Since the existence of the MX precludes A-Fallback , the A records MUST be referenced by the MX).
I guess I will need to be more explicate in the future and avoid unstated assumptions <g>.

Side question: Are you using term EXPLICATE intentional? because I think everyone is on sync using "Explicit" vs "Implicit". Implicit MX is the term used in 2821.

Anyway, 99% of the time I don't judge by POLICY. I used persistent protocol logic first, then use POLICY if there is particular reason whey one design over another may be deemed problematic. It could be an ethical engineering issue as well. i.e., I will never open an "Pandora Box" design. I wouldn't be able to sleep if I did.

That said, we have a IPv4 SMTP working model for the last 20+ years. By extension, IMV, any reasonable engineer would model IPv6 in the same way. Another way to say that is if you are attempting to fit IPv6 into the IPv4 world, then it would be common sense that IPv6 will behave with all the expectations of a IPv4 model. In other words, IPv6 should not alter IPv4 behavior.

That would be my justification from an engineering standpoint why an IPv6 client "may" be able to do a AAAA lookup, simply because an IPv4 client can do a A lookup. So this means, the IPv6 MTA does not have to have a MX record.

Now, is there a new policy reason or a new technical reason why we should extend the IPv4 "implicit MX" design to IPv6?

Maybe, maybe not. What would be the policy or technical reasons to not extend the same protocol logic we have not for IPv4 to IPv6?

Well, I did see some IPv6 field testing presentations that suggest that doing AAAA lookups are "Bad". Its not efficient and there are DNS servers that are yielding erroneous results.

Is that reason enough? Its interesting but I can't see that is reason enough. Not without more proof and confirmation. Maybe there are other factors involved here.

I did sense the feeling from several postings that some would not want AAAA lookups because of the predicted world of millions of small IPv6 devices on consumer machines, ones that are predicted to be sending out Email Reports or do mucho Call Home functionality. To me, thats fine, as long as the Return Path is valid - SMTP 2821 is not violated.

Is there a policy reason based on Security?

Well, I think the Crockers and Levines think so.


Hector Santos, CTO