On 03/06/2002 08:25:18 AM Michael Everson wrote: >That almost answers my first question. Does Devanagari glottal have >an inherent vowel? If it does, encode a new character.
That seems like a very good metric to consider, and I hadn't thought of it myself. I'd expect that this can be used syllable-initially rather than only finally, and so would have an inherent vowel but don't know that for certain. I've asked my contacts working in S. Asia for further info. >>(2) The second problem involves nukta (U+093C). In better-known languages, >>nukta can occur only on consonants, but for certain lesser-known >>languages, it can occur on vowels as well. Yet some implementations might >>not recognise a sequence like < consonant, vowel, nukta > as valid. For >>instance, I understand that if Uniscribe encountered such a sequence, it >>would assume you've left out a consonant immediately before the nukta, >>and it would display a dotted circle to indicate where a missing base >>character should go. > >So what would you suggest? A vocalic-nukta? I wouldn't like that. No, I wouldn't suggest anything different. The question is mainly intended to find out to what extent implementers are making assumptions that would present problems. >In >Cham, independent vowels can take dependent vowel signs. In >Devanagari, I guess that doesn't occur, but the Brahmic model >shouldn't be understood to preclude this behaviour. There's a general problem: writing systems of lesser-known languages sometimes involve behaviours that don't occur in the writing systems of better-known languages, but software implementations get designed based upon what is known, meaning the better-known writing systems only, and sometimes implementation incorporate constraints based upon what is exemplified in those better-known writing systems. E.g. there are Mon-Khmer languages spoken in Thailand that get written with Thai script but have many more vowel distinctions than Thai and so need to use combinations of combining marks not used in combination for Standard Thai, yet some important software implementations incorporate sequence constraints that treat these combinations as error conditions. >Um, that's AA, II, U, and O. What does the nukta make them sound like? I haven't any idea, myself. - Peter --------------------------------------------------------------------------- Peter Constable Non-Roman Script Initiative, SIL International 7500 W. Camp Wisdom Rd., Dallas, TX 75236, USA Tel: +1 972 708 7485 E-mail: <[EMAIL PROTECTED]>