Re: Minor is. It's not a pardigm change
This one is the one which I find the most compelling. In my mind, at the
moment the specification causes ambiguous behaviour because of the
implicit fallback to A records. Does 'connection refused' mean 'there is
no mail server here, so give up and don't come back' or 'the mail server
is down for maintenance/crashed/etc, so please do come back'? With the
implicit fallback to A, there's no way to tell, but with a required MX
record there is.
- mail queues get clogged up trying to send mail to hosts that don't
support SMTP. because "connection refused" is a temporary error to
SMTP, any mail sent to a host that doesn't support SMTP will sit in a
queue for days before bouncing or being discarded.
Spam being received by people won't be reduced by not allowing fallback
to AAAA, as the mail wouldn't get anywhere anyway, but it will be
sitting in queues on servers taking up resources unnecessarily.
I can see the need for this ambiguity in IPv4 now for backwards
compatibility reasons, and I can see why it was there originally for
backwards compatibility at that time, but now there is the prospect of
clearing things up and removing the ambiguity in the future, without
breaking backwards compatibility more than it would be anyway (IPv4 only
mail servers won't be able to talk to IPv6 only servers anyway).
*Now* is the time to do this. If the IPv6 mail standard is finalised
allowing fallback to AAAA, then the cat is out of the bag, and there's
no going back. So, unless there's a good technical reason (the only
reasons I've seen so far have essentially been 'ease of use' or
'political' reasons (the possibility of overly time consuming approval
processes to get another MX record added in the rare case where an
existing host without mail will need to have mail added)) I'm for not
allowing fallback to AAAA as it removes the 'connection refused' ambiguity.