A-label definition

[John C Klensin]

--On Monday, 23 June, 2008 22:01 -0400 Vint Cerf
<vint at google.com> wrote:

> seems to me that clarity might suggest a distinct RFC. the WG
> would   have to agree that such an RFC is in scope (I
> certainly believe it to   be on the grounds that we are having
> to look very carefully at how we   word our definitions as we
> introduce IDNs since we don't want to   create unplanned side
> problems with the introduction of the xn--   format at each
> level.

I still haven't been able to read every note in this thread,
but,   FWIW, this is the direction I think I was headed in with
my earlier note today.  After thinking about it a bit more, let
me suggest the following:

(1) We not spend energy tracing the derivation of LDH, hostname,
etc., rules and their relationships to IDNs.  We also avoid
spending any more energy than absolutely necessary on figuring
out what "will be alphabetic" meant in RFC 1123, how it relates
to some comments in RFC 1591, and how relevant either one is
today.

(2) Instead, let's move in the direction of getting a BCP
document together that makes recommendations about whatever the
IETF thinks is the best/ most desirable practice in this area.
IMO, we should be talking about best practice recommendations
because, for reasons that Mark and I have been discussing from
our rather different perspectives, the underlying DNS protocol
is able to handle a lot of strings that no sensible person would
use... these are really recommendations about how the DNS is
used in the protocols that call on it and hence about
registration policies, not the DNS protocol itself.  If we go
that route, the BCP should explicitly update/ quality the
language in 1034, 1123, 1591, etc.   

(3) If we do such a BCP, the "no strings that have hyphens in
positions 3 and 4 and are not A-labels" text should come out of
"Rationale".  It would either belong in the BCP (if consensus
exists) or not at all.   The IDNA2008 documents would presumably
still have something to say about strings that start in "xn--"
but that are not valid A-labels, at least wrt IDNA-aware
applications.

(4) I don't have a strong opinion about whether the BCP effort
should be on this WG's task list or not.  Either way, it would
clearly need to be carefully reviewed by the DNS Directorate and
appropriate DNS WGs and mailing lists, since its applicability
would be to DNS use and applications, not just to IDNs.

I think that document is more or less the one Mark was looking
for when he said "...RFC which consolidates all the changes to
hostnames: syntax, lengths, etc., into one document.... although
perhaps that turns it back into a standards track document that
also contains some recommendations.


I'd be happy to work with Mark and/or Frank to get such a
document together if that would help.  I don't think the text is
very complicated.  Getting consensus on its provisions probably
would be, but I'd consider it a success in the more difficult
areas if it could simply explain the tradeoffs to be considered
carefully s.t. the "best practice" would be to consider them.

Does that help?

     john