Now that I think about it, 2821bis-06 does require something more than
A/AAAA in order to operate a mail server: If a host only has an IPv6
address, 2821bis-06 section 5.1 requires the operator of that host to
form an arrangement with a dual-stack relay in order to send mail to
IMO, that's not an unreasonable requirement. I think it's perfectly
reasonable and almost unavoidable. I'm pointing it out to show that the
text already requires something more than getting an AAAA record and
setting up some software on the host itself.
How many people would be happy if 5.1 were to say this?
- if there is no MX record, but there is an A, then the sending
host MUST synthesise an MX
Has to be this way - any change to this is out of scope of this effort, not
to mention a very bad idea.
- if there is no MX record and no A, but there is an AAAA, then
the sending host MAY synthesise an MX
I have stated repeatedly that this is the one outcome that I view as completely
unacceptable, and since I have made this point multiple times I'm not going to
bother to repeat it.
IMO our choices here are between MUST and SHOULD NOT. I'm in favor of MUST but
I will (reluctantly) settle for SHOULD NOT if that's the consensus.
- even if a client doesn't use AAAA records for synthesising MX
records, it MAY use them for delivering mail once an MX has
I don't see how this warrants being a separate case to consider. You look up
MX, if you don't find you pretend you found one pointing at itself. You then
proceed to look up the A/AAAA records corresponding to the hosts on the MX
list. (If AAAA fallback is prohibited you also have to set a flag turning off
AAAA lookups, but this is no real hardship since you probably want to only do
AAAA lookups when IPv6 access is available.)
This permits people to set up IPv6-only mail servers without any MX or
relay, and it doesn't declare any deployed 2821-compliant code to be
illegal, and it doesn't tell anyone to assume that IPv6-only devices
are mail servers by default.
And in the process blocks our ability to move to draft without getting some
kind of exception. And I have to say I see no legitimate reason for the IESG to
grant such an excception.
Connectivity to/from AAAA-only mail servers will suck terribly for many
decades to come, but no 2821bis wordsmithing can avoid that. Only an
arrangement with a dual-stack relay can improve connectivity, and that
practically requires an MX RR or more.
This is certainly true for the Internet email service. But not all use of
standardized email protocols (including usage on the open Internet) is in
futherance of this service. It is entirely possible that at some point
an alternate service offering will emerge that's IPv6-only. I have no idea
how likely this is but it is not beyond the realm of possibility.