Frankly, I think Dave just enjoys arguing, but I see a point to clarify:
Dave Crocker <dhc(_at_)dcrocker(_dot_)net> wrote:
John Leslie wrote:
One could say, "Anything that eliminates the use of an A RR as a
substitute registration changes this flexibility."
I fat-fingered this, thinking "substitute for registration, but not
typing it. I do not believe Implied-MX is a "registration".
What this reformulation misses is that using an A* record means that
there is no incremental registration requirement.
... another of Dave's leaps of faith...
Since the A* record is there for other reasons, it means that being an
email server only requires running the software.
In fact, "being" an email server has never required anything beyond
running the software. Witness some extremely clue-challenged email
administrators we have all had the pleasure of dealing with...
So, what you are calling "substitution" is, in fact, requiring an
email-specific registration step that is not now present.
This is a difficult sentence to parse -- not made any easier, alas,
by my fat-fingered typing. But, as I read RFC2821, there clearly _is_
an email-specific registration step "present"; but there's also an
exception to it.
Dave seems to believe that the existence of an exception not only
dussolves any "requirement", but also requires us to make additional
exceptions whenever a "similar" situation arises. You missed your
calling as a Philadelphia lawyer, Dave. ;^)
In fact, it has already been documented over the last few days that
registering MXs can be problematic.
... as well as a number of other instances of the clue-challenged
running the railroad...
(Note, Dave is not saying these clue-challenged DNS administrators
would _actually_ prevent folks from exercising their right to squeeze
under the Implied-MX exception already in RFC2821.)
So the requirement "How should the intent to receive email for a
domain be signaled?" is, in fact, a very basic change to the core
Internet mail service model.
Funny, I thought it was a question.
There is currently no requirement to register intent. So to ask the
question is to look for a change to the model.
(Exasperation is creeping in...)
I _was_ careful to avoid the "evil R-word" there. An intent _can_
be "signaled" without "registration".
I'm fine with providing an ability to state intent. I'm not fine
with requiring it.
We're writing _standards_ here, Dave. Standards are _about_
requiring things. Making ever-more exceptions is something legisators
love to do, but it can never simplify the overall process.
I have expressed elsewhere my doubts whether folks coding email
server software are even going to follow the instructions we put in
2821bis. (I've even expressed doubt whether it's _possible_ to follow
the convoluted revision of Section 5 currently in 2821bis-09.)
Everybody here is nearing exhaustion. The "about a week" Tony
Hansen set for this discussion is drawing to a close. I don't envy
his task of calling consensus on this.
But while we're discussing the subject, I see nothing wrong with
raising the issue. Perhaps somebody might become convinced that
active registration is worth encouraging.
As long as the discussion is decoupled from the *bis round of
discussion, that's dandy.
I'd be happy to discuss it separately.
It's easy to get distracted.
John Leslie <john(_at_)jlc(_dot_)net>