On Jan 2, 2020, at 2:09 PM, Arnt Gulbrandsen
Running a v6-only host is stupid at this point,
I don’t share that assumption for multiple reasons. One is that the Internet
is a diverse place and there are lots of corner cases. It’s not the case that
every internet host needs to interact with hosts using ipv4. Another is that
v6 networks are likely to support some sort of NAT solution which permits v6
clients to contact v4 servers as long as those servers have dns names and
public ipv4 addresses.
In general I’d recommend that applications that can be IP version agnostic, be
so. As applied to SMTP I think that means accepting both IPv4 and IPv6 syntax,
but not actually caring what the peer’s address is.
perhaps slightly less stupid in five years, but even so I don't think the
next RFCs should feature an IPv4 requirement for address literals, and
nothing for the main parts of the protocols.
I would like for 5321bis to provide clear guidance to implementors rather than
having a practical necessity for them to implement protocol features that have
been removed from the standard. I expect that smtp and http will be the last
applications to not need ipv4 support somewhere. So I don’t think it’s helpful
to remove v4 address literals from the smtp protocol. However it might be
worth relaxing the syntax so that at least from a protocol perspective, smtp
ehlo doesn’t care what kind of argument is used.
ietf-smtp mailing list