ietf-smtp
[Top] [All Lists]

Re: Why implicit MX is a bad idea for IPv6

2008-04-05 15:54:41



On Sat, Apr 05, 2008 at 08:07:25PM +0200, Peter J. Holzer wrote:

IMPORTANT: a domain with no MX _and_ no A record, is not a maildomain.

This means an AAAA record is not an implicit MX record.

No, that doesn't follow at all. If all the MX records for a domain
contain hostnames with only AAAA records, the domain is just as
unreachable for an IPv4-only host as if it contained only an AAAA
record. And for a IPv6-only host it is just the other way around. 

It is the IPv6 host which is incompatible with existing standards and
it is the IPv6 host which will have to work around that.

And your believe is probably based on what you write next:

Many people arguing for the "only A records are implicit MX records"
side seem to ignore the fact that an MX record doesn't contain an
address on the right side. It contains a host name, which still needs to
be resolved to an address (via an A or AAAA lookup). The implicit MX
record isn't "synthesized from an A record", it is synthesized from the
domain name. There is no difference between the A lookup on the
hostnames in the real MX records and the one in the implicit MX records,
and neither should there be a difference in the AAAA lookups on those
hostnames, unless you want to follow the principle of most surprise.

2821 section 5:
"
   If
   no MX records are found, but an A RR is found, the A RR is treated as
   if it was associated with an implicit MX RR, with a preference of 0,
   pointing to that host.
"

which contradicts your statement and supports mine.


If that set is empty, this is an error condition.

IMPORTANT: a domain with no A record in its MX set, is in error.

No. It is merely unreachable from IPv4. This is *probably* an error, but
may be intentional.

Most definately, not probably.

2821 section 5:

"
   If MX records
   are present, but none of them are usable, this situation MUST be
   reported as an error.
"

And this is exactly what will happen when one or more MX records exist,
but none of them have A records.

Intentionally creating confusion just seems wrong.  I think correctly
programmed software will cope with "MX present but no A", but I wouldn't
be surprised if there's software out there which has difficulties with
such a case.

And don't forget the users.  The IPv4 host may just report "cannot deliver",
but it could equally well report "bad configuration". And some hosts could
keep retrying (which would be an error in itself, dunno).

Again: IPv6-only host setups will have to deal with the problems they
create.  That means setting up gateways and A records.

        Today this is a reasonable expectation.  Essentially IPv6
        only networks can still get some IPv4 addresses to operate
        their dual stack boxes with.

        NAT64 still requires a real IPv4 address.

        In the end it will be the operators of the IPv4 only legacy
        systems which will need to provide the support as there
        won't be any IPv4 addresses available to the IPv6 only nets
        to map through.

                e.g. NAT46 gateways.

        Mark
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark_Andrews(_at_)isc(_dot_)org

<Prev in Thread] Current Thread [Next in Thread>