Re: [idn] Re: Is space allowed in a hostname?

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

--On Wednesday, 10 July, 2002 01:07 +0900 Soobok Lee
<lsb@postel.co.kr> wrote:

>....
> By definition, NFKC = "compatibility decomposition" + NFC .
> This "compatibility decomposition" can be partitioned into
> groups of script-specific compatibility equivalences, and IMO,
> each group  has its own degree of appropriateness in the
> context of domain name. For example, only with NFKC,
> compatibility  ideographic character (Kc) can be mapped to its
> equivalent unified ideographic character (Ku). Kc and Ku share
> the same glyph, but often have different readings and origins. 
> That  mapping is OK. However, NFKC does not unify a certain
> pairs of TC/SC  which are purely in font-variant relations
> (like between italic/subscript A  and normal A) and share the
> same meanings and readings. When applied to hangul compat
> jamos (u+33??), NFKC makes errors. When NFKC maps circled "A"
> into bare "A", it may be useless or unnecessary in domain name
> context.

Ok.   We have established, if it wasn't obvious already, that
NFKC is an easy target.   That doesn't make it "wrong", it just
means that, in the tradeoff decisions that were made, some of
the consequences may be unfortunate for particular applications
and/or languages.  

But being able to explain what is wrong with it when viewed from
particular perspectives does not, ultimately, help.  There are
two things that are much more important:

(i) Demonstrating that there is a variant of NFKC (or something
else that does the same job) that is better across the entire
range of scripts and languages, not just a scripts or two or a
handful of characters.

(ii) Getting that variant reviewed and approved by a recognized
standards body with global scope and competence to do this type
of evaluation.  My guess is that the IETF is no closer to taking
on that role than it was a half-dozen years ago, i.e., no chance
at all (and for the same good reasons).

Absent such a proposal, in written form, and significant motion
toward approval by such a standards body, it is nearly a waste
of time to complain about NFKC: "no alternative" unfortunately
outvotes all other possible answers.

> Ideal IDN name normalization was carved to fit in NFKC.
> "if you sleep in a bed which is shorter than you, you should
> bend your  legs every  night ?  or  buy new bed ?"
> Stringprep/nameprep teach us to bend our legs.  :-)

Bending is far better than the version in which your legs are
simply cut off to fit.

regards,
    john