multi-level naming (was: RE: emoji (was Re: I-D Action: draft-klensin-idna-rfc5891bis-00.txt))

[John C Klensin]

There seem to be several discussions on some sort of two-layer
system going on in parallel.  Perhaps it is time to start
thinking about consolidating them into a mailing list and/or
future BOF.

However, those who want to go down that path should be aware
that we tried it before, at more or less the same time IDNA2003
was maturing as a spec, and motivated partially by strong
suspicions that IDNs were going to get us into very nearly the
mess we are in now, that the DNS couldn't handle the matching
issues that IDNs and user expectations seemed to call for, and
even an early version of the branding issues (then in the form
of "[DNS] keywords".

It has been long enough that I had forgotten that we actually
had a WG (at least the tracker says we did) and have no memory
of what is in a draft I apparently wrote, but...

Circumstances have obviously changed since the WG was shut down
(end of 2003 according to the tracker) and I hope we have
learned some things, but, to lower the risk of reinventing the
wheel, see

https://datatracker.ietf.org/wg/irnss/about/
https://datatracker.ietf.org/doc/draft-klensin-dns-search/

The date on the last version of the latter seems to be after the
WG was abandoned, so it might constitute a summary of how far
the WG's thinking got.  Or not... again, my memory of that
effort is shot.  Glancing at it now, it at least contains a list
of topics to be addressed.

best,
   john

--On Thursday, March 16, 2017 21:05 +0000 Shawn Steele
<Shawn.Steele at microsoft.com> wrote:

> A few of us were thinking maybe identifiers almost need (might
> be too strong a word) two layers... One to provide a fairly
> strict version of providing "identifiers" in a way that tries
> to reduce confusion, and another layer that provides "friendly
> labels" that help people get to those identifiers in a way
> that makes them feel good.
> 
> Most (AFAICT) of the more interesting uses of IDN already
> resolve to a name that would fit in that "stricter identifier"
> bucket, so, in practice, we kinda already have two layers.
> The "marketing thought it would be good if this linked to us"
> which goes to the "this is the label that doesn't scare the IT
> department."
> 
> Regardless on where you draw the line of what characters are
> appropriate, this is already happening somewhat naturally,
> especially when it's "easy".  Eg: an umlauted domain resolving
> to a pure-ascii variant spelling.  Of course that's tougher
> for other languages, but the same idea seems to happen a lot.