From: Asmus Freytag (asmusf@ix.netcom.com)
Date: Fri Jun 04 2004 - 19:43:12 CDT
At 02:21 PM 6/4/2004, Peter Kirk wrote:
>>>>>There is no consensus that this Phoenician proposal is necessary......
>
>I am revisiting this one because I realise now that Ken has been somewhat
>economical with the truth here. There ARE cases in which entire alphabets
>have been given compatibility decompositions to other alphabets.
There are alphabets and alphabets.
>For example there are the Mathematical Alphanumeric Symbols, the Enclosed
>Alphanumerics, and the Fullwidth and Halfwidth Forms, as well as
>superscripts, subscripts, modifier letters etc. These symbols
Most of them are symbol sets that are graphically derived from alphabets.
They don't function like alphabets and you don't see entire text runs in
one of these sets, but a character at a time.
>have these compatibility decompositions because they are not considered to
>form a separate script, but rather to be glyph variants of characters in
>Latin, Greek, Katakana etc script. Do these compatibility decompositions
>cause technical difficulties?
Yes, they do, If people apply them. Compatibility decompositions modify the
semantics of a text (a known feature of them) so you must limit their use
to situations where that is the intended effect or you will disappoint your
users.
>>Compatibility decompositions directly impact normalization.
>
>Of course. And the point of suggesting compatibility decomposition here is
>precisely so that compatibility normalisation, as well as default
>collation, folds together Phoenician and Hebrew variant glyphs of the same
>script.
No. Compatibility decompositions are a blunt instrument as it stands. Read
UTR#30 Character Foldings, to see what I mean.
>>Cross-script equivalencing is done by transliteration algorithms,
>>not by normalization algorithms.
>
>This begs the question.
No, I think Ken's been very clear here.
You continue "... just as one would not describe as transliteration
representation in Times New Roman of a Latin script text in mediaeval
handwriting or in Fraktur", which in fact is incorrect. Fraktur uses the
long s which needs to be transliterated to normal s.
>If what I have suggested is ridiculous, so is what the UTC has already
>defined for Mathematical Alphanumeric Symbols.
It's not, because it's not the same issue. Symbol sets and alphabets are
different. And trying to re-open and re-argue this repeatedly on this list
doesn't improve the analogy.
A./
This archive was generated by hypermail 2.1.5 : Fri Jun 04 2004 - 19:44:45 CDT