Re: [idn] Intro to my I-D

[Eric Brunner-Williams in Portland Maine <brunner@nic-naa.net>]

Ben,

Now that Deng and I have had the chance to develop your ideas in this series
of letters, here are our conclusions:

> 1.  Registrants given the freedom/choice to point Tradition CDN to
> tradition website and Simplifed CDN to simplified website.

Restated in the sense of [30]:

	[30bis] Within a single equivalence class ["zone"],
	the registrant ["zone manager"] MUST be able to define
	traditional/simplified Chinese equivalence rules
	that suit the purpose of the registrant [zone].

	Such defined equivalences MUST [MUST NOT] remove equivalences
	that are assumed by (old or local-rule-ignorant) caches.

That is the scope issue. To paraphrase Harald Alvestrand, this creates
a possibility of "maximal surprise".

As Deng pointed out, your #1 has a geometric scailing property, with
any character string of N Chinese characters having 2^^N zone file
representations, when the character string is a name label. As a UTF-8
Chinese string may be up to 21 three-octet characters long, and each
character may have one (or more!) equivalences, this yields 2,097,152
possible distinct "names" reserved from one SC or TC IDN registration.
Slightly smaller numbers are used in the ACE discussions, so here I use
18.

A zone with bi-modal length distribution of Chinese business names, with
clusters at 8 characters and 18 characters, and 1M distinct registrants,
would have a zone file with 130,000,000,000 entries.

	.5M x 2^^8 + .5M x 2^^18 (dual "short" and "maximal" form)
	1.28e+08   + 1.31e+11 (xcalc numbers)

This appears to be "very large". You are aware of this, neh?

> 2.  A need for Tradition CDN to Simplified CDN conversion.

This goal, which you state your system recognizes with equal importance as
for goal #1, appears to be depricated (impossible) by the requirements of
goal #1.

Further, as I mentioned previously, and above in the stilted language of
scope for rules, the "conversion" is per-registrant, and effects a private
system of translating names to addresses.

What is the key value for users "opting out" of SC/TC equivalences?

Eric