Re: [idn] NSI Multilingual Testbed Information (fwd)

[Frank Ernens <fgernens@enternet.com.au>]

Bill Manning wrote:

> 	There are several ways to provide "multilingual" capabilities
> 	and many TLD registrars have selected a methodology. Why
> 	pick on this one when NU, CN, JP, KR, et.al. have followed
> 	similar paths.  Just because a method is not an IETF
> 	"approved technical specification" does not mean that it
> 	is not defined.

Unless the service being provided is just reservation of the names,
end users must be using clients such as email programs and web
browsers to access these names. If the non-Roman characters are
any use at all, the user will want to type them in where a standard
hostname goes. Are these people using patched/custom clients? How
widely used are these names, or are they just experimental? Is
there any support in the major commercial browsers?

There ought to be a single algorithm to transform a Unicode
name string into an ACE one (if we're using ACE, which I assume
these schemes do), not a plethora. Otherwise, every resolver
will have to handle every algorithm *forever*, assuming the
"multilingual" names are meant to be permanent. The increased
complexity and decreased performance is something everyone else
not using these schemes has every right to complain about.

Keith Moore wrote

> is it just me, or does anybody else get the impression that NSI is 
> way, way, way out of line here?

It certainly reads, complete with .net domain, as if it's something
official. Are the customers and investors aware that it isn't? Should
the IAB say something publicly or to the regulators?

This is not rocket science - I reckon it would take less than 2 hours
to stitch a Unicode library to a working DNS server which used
hashing. The entire question is one of what to do, not implementation,
and therefore anyone "jumping the gun" is not offering anything
of value but merely trying to gain an unfair advantage.