Re: language (or script?) tagging in draft-obispo-epp-idn-00.txt

On Thu, Dec 22, 2011 at 05:53:31PM +0000, Keith Gaughan wrote:
> Instead, 'writing system'[1] would be better: so long as the writing system
> contains no homographs[2], there's no issue.

Alas, most alphabetic languages are full of homographs.  But I bet
that isn't what you mean, anyway; homograph turns out to have an
established meaning in linguistics, and it isn't "two letters that
look the same".  In the ICANN VIP discussions we settled on
"homoglyph" as a possibly useful alternative.  Cary Karp has extensive
opinions about this.

> [1] By which I mean, all the languages that use the Latin alphabet could be
>     said to share a common script and writing system, which are both the same
>     thing. It's not an ideal term, I know.

Indeed, it's wrong, as a consideration of the fall-back handling of
o-with-diaeresis in Swedish and German illustrates.  German considers
ö to be an o with an accent on it.  In Swedish, it's a separate
letter.  That's why German orthography can fall back to "oe" but
Swedish doesn't.

> [3] I can't recall whether Traditional and Simplified Han characters occupy
>     different code points in general, or the same ones.

They're different code points, and different Abstract Characters in
Unicode terms; but there are usually pairs (with some nasty corner
cases) of the same Conceptual Characters in TC and SC.  Also, since
they're all part of Unihan, they're all "the same script". 

Slightly worse, Japanese uses the same code points, and they have a
_different_ pair of Old Form/New Form changes that were undertaken,
but they don't exactly regard them as the same Conceptual Characters.
This is because, I'm told by our friends at JPRS, they consider domain
names to be proper nouns and therefore always to be spelled exactly
one way.

Best regards,

A

-- 
Andrew Sullivan
<[email protected]>
_______________________________________________
provreg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/provreg