Re: [idn] The layers and character handling in DNS

[John C Klensin <klensin@jck.com>]

Paul,

I think this leads to results that violate the law of least
astonishment.

Suppose I'm using local character code X to represent script Z
and so is Keith.  My local mapping is Unicode<-F1(X,Z) and
Keith's is Unicode<-F2(X,Z), and they are different.  Let's
assume that each is consistently reversible (although that is
another problem, both with this issue and, I suspect, with
Nameprep).
 
So I find an IDN domain on the network, and, by the time its name
emerges from my application, it is in X, having been passed
through F1'(Unicode,Z).  I pass X to Keith out of band (e.g., in
an email message with "text/plain; charset=X").  Now Keith looks
it up, applying F2(X,Z) before Nameprep.  And, behold, he either
gets 'nodomain' or, worse, the wrong stuff.

I think consistent behavior is important.  And I don't see how
one gets that without either native implementations of 10646
(i.e., no local character sets) or with standardized/ unambiguous
mappings between each local set and 10646.

Of course, we could try to make Nameprep complicated enough to
smooth over all such details, s.t., Nameprep(F1(X,Z)))=
Nameprep(F2(X,Z)) for all Fn.   Seems unlikely somehow.

     john



--On Thursday, 15 February, 2001 12:33 -0800 "Paul Hoffman / IMC"
<phoffman@imc.org> wrote:

> At 3:08 PM -0500 2/15/01, Keith Moore wrote:
>> whether you can effectively *input* IDNs without platform
>> support for Unicode is a different question.
> 
> Correct. Most "platforms" have support for some character sets,
> each of which is a subset of the Unicode (ISO/IEC 10646)
> character repertoire. Because of this, the input mechanism can
> be used with IDN. The "platform" decides the mapping between
> the character sets and Unicode.
> 
> The difference here and where this thread was going is that the
> mapping is done by the application (probably through "the
> platform"), *not* by the protocol itself.
> 
> --Paul Hoffman, Director
> --Internet Mail Consortium
>