Re: [idn] Re: IDNA: is the specification proper, adequate, andcomplete? (was: Re: I-D ACTION:draft-ietf-idn-idna-08.txt)

[Patrik Fältström <paf@cisco.com>]

--On 2002-06-17 13.16 +0200 Simon Josefsson <simon+idn@josefsson.org> wrote:

> This seem to argue against the current design of IDNA.

I don't agree.

It is an example of IDNA solving the problem of using identifiers with
non-ascii codepoints.

IDNA does _not_ (and I think that has been clear all the time from us
writing the spec) solve the problem with use of "words" with non-ascii
codepoints.

As Vint wrote, DNS is extremely bad at the functionality needed for word
matching. Issues like locale dependent matching rules, ambigous result
sets, more interactivity all are non-existing in the lookup system DNS
defines.

Regarding the "normalization for resolving ambiguities", I think we talk
about different kind of ambiguities. In the case of normalization we only
take care of the ambigous representations that, according to UTC, exists in
the design of the Unicode Character Set. If you do word matching, you get
different (often-locale-dependent) ambiguities, part from of course
context-dependent ambiguities which exists because two very different
things can be named exactly the same.

     paf