Re: [idn] Intro to my I-D

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

[I've added idn@ops.ietf.org back to the cc'd list.
 Harald, please read and reconsider the "principle of least surprise".
 Thx.]

Ben,

What you've proposed is that for some collection of code-points, and for
some equivalency rules (whether zone-scoped or global-scoped), allocation
of one code-point in the collection results in pre-allocation (reservation)
of all code-points in the collection.

Additionally, you've proposed the registrant (recipient of the allocated
code-point equivalency class) may then modify the equivalency rule which
created the class, and assert distinct semantics for distinct elements of
the original (intact), now fragmented class.

To help me understand your proposal, I'll consider ASCII strings and the
Pan-Algonquin "OU" character (U+0222) and (U+0223). A locally constructed
French character, not present in 17th century French. I know that most of
the North Americans and Europeans will find this choice strange, but more
accessible, as "Indian French", than an ideogramatic example, and I know
that my friends at CNNIC will speak to the Chinese language specifics.

Assertion: (zone scope)

	In the Algonquin-preferred zone(s), e.g., abenaki.dst.me.us,
	the following code-points and code-point sequences are
	equivalent:
		U+0070		"8"	when in an alpha-string
		U+0222		"8"	LATIN CAPITAL LETTER OU
		U+0223		"8"	LATIN SMALL LETTER OU
		U+0117,U+0125	"OU"
		U+O117,U+0165	"Ou"
		U+0157,U+0165	"ou"
		U+0127		"W"
		U+0167		"w"

Under your proposal, in my zone (abenaki.dst.me.us.), with ASCII case folding
(applied to the two Latin Extended-B code points), the registrant for
"ki8na.abenaki.dst.me.us" ("ki8na" means ourselves or nous-memes), would be
allocated or reserved:

		ki{set of 4 code-point sequences}na.abenaki.dst.me.us 

The registrant could then be able to associate between 0 and eight unique
ip addresses to these 4 allocated domain names in my zone, e.g.,

	ki(U+0070)na.abenaki.dst.me.us and 1.2.3.1
	ki(U+0222)na.abenaki.dst.me.us and 1.2.3.2
	ki(U+0223)na.abenaki.dst.me.us and 1.2.3.2
	ki(U+0117,U+0125)na.abenaki.dst.me.us and 1.2.3.3
	ki(U+O117,U+0165)na.abenaki.dst.me.us and 1.2.3.3
	ki(U+0157,U+0165)na.abenaki.dst.me.us and 1.2.3.3
	ki(U+0127)na.abenaki.dst.me.us and 1.2.3.4
	ki(U+0167)na.abenaki.dst.me.us and 1.2.3.4

Without case folding there could be 8 distinct addresses rather than just 4.

This would be very surprising to a modern literate North-Eastern Indian, and
to scholars of the Contact Period and French and Indian literatures of the
17th and 18th centuries. As the abenaki.dst.me.us zone manager, I don't think
I would allow it. I would expect the equivalency rule (above) to be adopted
by the {penobscot,passamquoddy,maliseet,micmac}.dst.me.us zone managers, and
by other "Algonquin-aware" NSN.US zone managers. I would not expect these to
be adopted by the dst.me.us or superior zone managers. I would expect the same
rule and scope to exist in the .ca zone and its dependent zones.

My preference is that ki{set of 4 code-point sequences}na.abenaki.dst.me.us
all map to a single internet address.

In the Chinese language area, which includes North America as well as China,
your "registrant election" proposal appears to have the same surprising set
of consequences as one registrant following one ideosyncratic convention for
"ki8na" and another registrant, possibly the same one, following a different
ideosyncratic convention for "ni8na" (we and you or nous et vous), and this
only gets worse as we consider strings such as k8sk8ranmsh8dans8 (sift).

Just what makes your proposal superior?

Eric

P.S. The advantage of [30], formerly "in" the IDN WG requirements draft, was
that it made zone-specific semantics possible. Writing in modern Abenaki (or
any "8" using French-originating modern script) using ASCII equivalency rules
presents a problem for 0x70 to 0x127,0x167 mapping (simplest case).