From: Charles Lindsey (chl@clw.cs.man.ac.uk)
Date: Mon Jul 02 2001 - 06:17:25 CDT
In <20010629152956.O8140@demon.net> "Clive D.W. Feather" <clive@demon.net> writes:
>Charles Lindsey said:
>>> Do you have a reason for making compatibility characters a MUST NOT rather
>>> than allowing hierarchies to override ? The basis of my approach was to use
>>> the latter rule for anything where we could get away with it.
>> It just removes another source of confusion, and since Unicode have
>> provided it as a "feature", we may as well build on their work.
>How does this differ from the ban on titlecase characters ? That's the bit
>I have trouble with.
I think because the things that would be removed by insisting on NFKC
normalization are an order of magniturde more ugly and confusing than
titlecase characters. And also that the Unicode people have provided this
nice tool, which will be widely used in various places, which leverages so
much benefit from an easy-to-express restriction. Note that I am assuming
that the Unicode people have done it properly, so that no language likely
to be used on Usenet will be impacted.
>> Would anyone object to "glyph" (meaning the same as "locale independent
>> grapheme"), or has someone else somewhere defined that in a confusing
>> manner?
>I'm not sure that has a good meaning or the meaning we want (remember that
>in some languages several characters combine to form a single glyph). Why
>not just define the count in terms of the appropriate entity in the syntax,
>which is "component-item" ?
Yes, but what I had in mind was to replace the term "component-item" by
"component-glyph" or "component-grapheme" or "component-other-foobar" that
would be more helpful in leading the reader's perceptions in the right
direction. But would that help or hinder, and what is the write "word" to
use?
>> HOWEVER, I think I would now prefer to make NFKC the "main" definition,
>> and put the "compatibilty decomposition" and the "NFKC-NO" as the bulleted
>> alternatives. Would you object to that?
>Yes, I would.
>Previous discussion on this topic showed me that the "must be invariant"
>rule was rather less well understood. In particular, people were taking it
>to require normalization to be carried out.
OK. Point taken.
-- Charles H. Lindsey ---------At Home, doing my own thing------------------------ Tel: +44 161 436 6131 Fax: +44 161 436 6133 Web: http://www.cs.man.ac.uk/~chl Email: chl@clw.cs.man.ac.uk Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K. PGP: 2C15F1A9 Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5