Re: [idn] hostname history hell

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

--On Tuesday, 20 November, 2001 15:48 -0800 Tim Langdell
<tim@xtns.net> wrote:

>>  We don't _need_ them for identifiers.  Some of them will,
>> sooner
>> or later, run up against a legitimate command language or
>> cause "interesting" lexical parsing problems (even if they
>> don't cause problems in today's URI syntax definition).  High
>> risks, marginal benefit.
>>...
> I am in substantial agreement with you John except I think on
> this issue of symbols and drawing characters. Our market
> research shows that there is going to be a high degree of
> demand for such characters. Further, with the wide range of
> scripts and characters supported by Unicode, for many it will
> be a rather academic hair-splitting which is a language
> script/character and which is technically a symbol or drawing
> character.

Since you introduced the term "market research" into this
previously high-minded discussion ( :-( ), have you thought
about what you would put into a whois-like table with these
things?  How you would feel about being put into the middle of a
discussion of whether a name starting in a left-wizzlepop was
too similar to one starting in a left-popplewiz? I may have
missed something, but my impression is that the UTC TRs that
describe canonicalization don't do much for drawing characters:
do you propose that the IETF do that work?  Or do you think it
is unnecessary?   

And so on.  The point here is to make a case for "more trouble
than they are worth" -- people would _like_ lots of things from
the DNS, including things we don't know how to do (which is why
many of us are working on "search layer" models); I'm suggesting
that a little drawing of the lines short of the conceivable
technical limits might be to almost everyone's advantage.

     john