From: Richard Wordingham (richard.wordingham@ntlworld.com)
Date: Wed Sep 07 2005 - 17:36:24 CDT
Kent Karlsson wrote:
> Peter Constable wrote:
>
>> > It seems to me that you are going beyond what is currently spelled
>> > out by Unicode. The discussion about the joiners does not mention
>> > how they interact with vowel signs.
>>
>> True; it is silent on that issue. For better or worse, in the context
> of
>> that silence we made an implementation decision. That decision is not
>> inconsistent with what the Standard specifies.
>
> It still appears to me that requesting virama to be visible is mostly
> orthogonal to requesting a reordrant vowel to "reorder less".
Isn't that because you think that a virama codepoint between consonants in a
script encoded using the 'virama model' has got something to do with a
real-world virama? Would you think the same if the script's encoding used
the coeng model or had a separate encoding for either (1) non-final
consonants of a cluster (no examples, I believe) or (2) non-initial
consonants of a cluster (e.g. Tibetan)?
I see it quite differently because I think it is a matter of what
orthographic syllables the sequence of vowelless consonants (as you see it)
is split into - in the output if there is licence to to do it differently to
what is specified in the sequence of codepoints as the ideal form. I
suppose one may need to cater for a dependency of vowel placement on the
requested grouping of consonants into orthographic syllables as opposed to
the delivered grouping. Does this exist in any real language? I quote from
the first post Gautam Sengupta made to indic@unicode.org on 7 September 2005
GMT:
> - p127 entry for "peer VI", last word contains dda + halant + sa +
> reph
> - p142, entry for "range, N", 3rd word contains matra i + nga +
> halant +
> k.ta
There are no such words in Hindi, contemporary or old. No native
speaker will recognise those sequences as "well-formed". A halant
after a vowel sign and a reph after a halant are ill-formed.
(end quote)
The final sentence is language and script dependent - Burmese uses a
combination of vowel and virama for one vowel. However, it seems that not
even CrrC (in what language?) will give <halant, repha>.
> Further, I think this is a "behavioural" issue, rather than just an
> implementation issue. There should be a common (read standard)
> way of dealing with this, at the character sequence level, independent
> of how the implementations are built.
Agreed.
Richard.
This archive was generated by hypermail 2.1.5 : Wed Sep 07 2005 - 17:38:53 CDT