On 4/7/08 at 6:54 PM -0400, Hector Santos wrote:
So by virtue of the fact that 974 was not specific, in hindsight you
are suggesting it left it open-ended for any future record type
associated with the domain?
Yes, though not on purpose. I'm actually coming at this from the
opposite end of things than I believe you are:
When was 974 written? 1986?
When was IPv6 invented? 1991?
When was AAAA invented? 1995?
And let's remember to add that 2821 was added in 2001.
Lots of things had to change when IPv6 was introduced. A new AAAA
record was needed to represent addresses in the DNS; a new EPRT
command was needed in FTP. But there were lots of things that didn't
change. We didn't need a new UDP header. We didn't need a new MAIL
FROM command in SMTP. We didn't need new Telnet options negotiation.
And what I'm claiming is that (prior to 2001), we didn't need a new
MX record handling algorithm, as per the one in 974. The one in 974
was perfectly applicable to IPv6: If you get back an empty MX list,
you make the 0 preference one as described and look up the address.
Yes, that address lookup is different than it was in IPv4, but (as
the examples above show) that doesn't mean that anything else in the
protocol needs to change. And in fact, a well-designed protocol
shouldn't change at every layer just because a single layer has
The MX rule of 974 is perfectly reasonable to apply to IPv4 and IPv6.
I think that's what we should have 2821bis do.
Pete Resnick <http://www.qualcomm.com/~presnick/>
Qualcomm Incorporated - Direct phone: (858)651-4478, Fax: (858)651-1102