On Sun, Mar 30, 2008 at 10:51:25AM -0700, Russ Allbery wrote:
Alex van den Bogaerdt <alex(_at_)ergens(_dot_)op(_dot_)het(_dot_)net> writes:
Then: people didn't know MX records and thus didn't use them. Or maybe
they did already know about MX records but did not yet support them.
Some fallback mechanism was necessary to bridge the gap between pre-MX
and post-MX time, or else MX didn't stand a chance.
Now: people do know MX records and do use them, except for a couple
of spammers maybe. There is no need anymore to design a fallback.
I think you should collect some data on this before blithely making such
assertions. All my available anecdotes contradict your assumption so
strongly that I find it very dubious.
For example, here at Stanford, there are a ton of departmental servers
that accept mail for some reason, and almost none of them have MX records.
People only add MX records when they need to direct mail addressed to one
system to a different one instead.
OK, let me rephrase.
Back then, almost a catch-22 situation would exist without the fall back.
Nowadays, using MX records or not is a choice. People know they exist,
how they do their stuff. Should a fall back not exist, MX records can
be added quickly without much effort. Not publishing an MX record is
a choice of using the implicit MX -or- the host does not receive mail.
In general, you should count on human laziness. If mail works without
adding an MX record, people aren't going to add one. If nothing breaks
when the MX record is missing, no one will notice it is missing.
Then the opposite also holds. As soon as people find out mail doesn't
work as soon as they are on IPv6 only, they will add an MX record.
I am not arguing that the implicit MX should go for IPv4(*). I am just
arguing that it shouldn't appear for IPv6.
My gut feeling: In IPv6 there will be a lot more hosts, and a
significantly smaller percentage of those hosts will receive mail.
Saving in DNS lookups will outweight the inconvenience of publishing
MX records for those "few" hosts which require it and wouldn't with
an implicit MX.
(*) I know, IPv4 isn't its official name.