Re: [idn] IDNRA comments

["Maynard Kang" <maynard@pobox.org.sg>]

Hi,

> My reference was meant as an example, in order to describe a TYPE of
> transformation, and was not meant to recommend the specific MIME
> content-transfer-encoding techniques.

I must have read too fast. My apologies for the misunderstanding.

> On the other hand, these techniques very much DO preserve the uniqueness
of
> a string, including when that string is a domain name, since the MIME
> content-transfer-encodings must be entirely idempotent and reversible.

I'm afraid I have to disagree. The transformation itself may be idempotent
and reversible, but the fact that any arbitrary character set may be used in
MIME content-transfer-encodings makes the domain name (not the string)
non-unique.

For example, the same abstract characters (zhongguo.gongsi in chinese, which
means china.com), ideographically speaking, can be represented in at least
three different ASCII-encoded strings using QP:

=?GB2312?Q?=D6=D0=B9=FA.=B9=AB=CB=BE?=
=?BIG5?Q?=A4=A4=B0=EA.=A4=BD=A5q?=
=?UTF8?Q?=E4=B8=AD=E5=9B=BD.=E5=85=AC=E5=8F=B8?=

Perhaps I should clarify, I agree that the *transformation* results in a
unique string. But not a unique domain name in the case of multiple CCS
used. The transformation is fine, but for use in IDNs the CCS should be
restricted to a single, universal one such as ISO10646, like in the case of
RACE and UTF-5.

-maynard