Lookup for reserved LDH labels

[John C Klensin]

--On Wednesday, 07 November, 2012 13:01 +0100 Simon Josefsson
<simon at josefsson.org> wrote:

>...
> After considering that IDNA2003 did not fail, and theere seems
> to be real domains registered like that, let me take back that
> statement.
 
> I don't have any opinion on what the intention behind IDNA2008
> was, but given that IDNA2003 permitted the domain, there are
> registered domains like that out there, and the IDNA2008
> lookup model is to trust DNS, and IDNA2008 has specification
> unclarity/bugs in this area, I'm somewhat inclined to feel
> that lookup for that address should work.

Simon,

<Insert my disclaimer about opinions as to why things were done
by reference>

Remember that IDNA2008 explicitly, and after long discussion,
forbade lookup of XN-- strings that were not A-labels in spite
of understanding that there were native character strings out
there and registered that did not conform to the U-label rules.
Some of those strings were explicitly forbidden by IDNA2008
(e.g., ones that contained DISALLOWED characters), others were
invalid under IDNA2003 but still looked up (e.g., those
containing code points that were unassigned in Unicode 3.2).  

I believe it was also the WG's explicit intention to invalidate
RACE labels and provide disincentives to their further use.

Personally, I find it hard to reason that the WG intended to
forbid/ break labels that were registered/delegated and
perfectly valid under IDNA2003 but still intended to allow
lookup of labels associated with other prefixes (or, for that
matter, labels containing non-ASCII characters but coded
directly in UTF-8) to be looked up in the public DNS.  The
practical impact of permitting those other forms is an
invitation to interoperability, or at least performance,
problems as a client has to experiment with multiple encodings,
conventions, or strategies in the hope of finding a name and
then having to figure out which set of RRs are intended if more
than one name is found.

To paraphrase a conversation in a WG meeting earlier this week,
if a special label convention is needed for another purpose, it
is probably better to look at variations on the "_" scheme.
Because it uses a subdomain structure, the namespace/ codespace
is much less limited (thanks, Martin, for reminding us about
that part of the puzzle).  "__zz--CCCCC" is not, by any
interpretation, a putative A-label.

best,
    john

p.s. thanks, Martin, for the rest of your note.  I completely
agree.


> 
> /Simon
> _______________________________________________
> Idna-update mailing list
> Idna-update at alvestrand.no
> http://www.alvestrand.no/mailman/listinfo/idna-update