> -----Original Message-----
> From: James Seng [mailto:jseng@pobox.org.sg]
...
> Special note: I added
>
> The protocol must only permit characters in labels which is a valid
> characters in a language. In other word, control sequence, symbols,
> punctuation, etc must not be used in label. The exception to this
> requirement is '-' (U+002D).
>
> Which Paul comment:
> [[[JAMES: There was not agreement to the above paragraph, and I think
> that there will be violent disagreement. For example, there are many
> scripts that have "punctuation" that is used just as often as "-" is
> used in English. Further, I see no technical reason to exclude symbols
> from IDN. I propose that you remove the paragraph and bring it up on
> the list.]]]
Suggested replacement text:
"Registrars should only be allowed to register names that consist
only of characters that are used in words, according to some
orthography. Thus punctuation (except in-word punctuation, like
hyphen), symbols, control characters, or format control characters,
should not be used in a name. Period, including full-width period,
is reserved for separating parts of a domain name. The *amended*
ISO/IEC TR 10176 Annex A (Recommended extended repertoire for
user-defined identifiers) should be considered when selecting
the IDN character repertoire."
Rationale:
I see no reason to make such requirements on the protocol itself,
just the registration process. And these kind of recommendations
are on the 'should' level, not the 'shall' level, especially
in the requirements document, perhaps also later. This kind of
recommendations are there to retain the word-like nature of
domain names, not for any strong technical reason, other than
that symbols and punctuation may be used for other purposes in
close proximity to a domain name, as some punctuation already is.
Kind regards
/kent k