--On Monday, 07 April, 2008 16:00 -0400 Hector Santos
<hsantos(_at_)santronics(_dot_)com> wrote:
John C Klensin wrote:
...
I think the chart, summary I provided shows that the question
we have attempts to address how a IPv6 system should behavior.
Not a IPv4 system.
It is the IPv6 system that needs to decide where it should
look up A or AAAA records. Not IPv4 systems.
Unless you don't think so, I don't think 2821bis is in any
position to begin describing how IPv6 should behave as far as
what records and/or order of lookups it should do. A first,
AAAA second or vice versa, information that both RFC 3484 and
also 4083 attempt to address for IPv6 systems or hybrids. Not
pure IPV4 systems and this is clearly described in both
documents. IPv4 only systems will behave only one way. This
is independent of a Dual Stack Node. Not Dual Stack SMTP
Application
In short, IMV, 2821bis is still about IPv4 and I think the
only thing that we can even mention is to codify the fact that
an incoming IPv6 still needs to be compatible with a 2821 IPv4
SMTP system.
Hector, I read all the way to your last paragraph and was
preparing to write an "I agree" note. Than I read that
paragraph and realized I had misinterpreted everything that you
had written earlier.
2821bis is intended to apply to both IPv4 and IPv6. We really
have no practical model for specifying an application,
especially an application that runs over TCP, that runs
exclusively over (TCP over) IPv4. It is not an accident that
2821 specifies a syntax for IPv6 address literals... and, unlike
the change in wording about MX records from 974, that change was
quite explicitly discussed during the DRUMS work. It is very
clear that the fate and handling of AAAA records is irrelevant
to an IPv4-only implementation and almost equally so to a
dual-stack implementation running in an environment with no IPv6
connectivity. But that doesn't make 2821bis "about IPv4" or
imply that there will be an SMTPv6 somewhere down the line.
john