James E. Agenbroad wrote:
>
> Whether a vowel or vowel sign can have a nukta I do not know.
As a sidepoint (not really relevant, but it may matter), ISCII-91
does use the combination of vowel with nukta, in both standalone
and vowel sign form.
short ri + nukta gives long rii
short i + nukta gives short vocalic li (and issues a collation problem)
long ii + nukta gives the theoric long vocalic lii
> 6. Two adjacent halants or vowel signs are probably invalid.
Again, this was "reused" by ISCII-91: the combination halant+halant
means "explicit halant", much like the use of zwnj in Unicode.
> 4. In a conjunct consonant only the last consonant can have these
> 'various signs' after it (possibly with nukta or vowel sign (or
> halant?) between the consonant and one of the 'various signs.
Well, there is a problematic point here: with the sandhi of n with l.
The PaNini rules ("torli" IIRC) asserts that it should be rendered
as nasalized l followed by normal l with the associated vowel (whatever).
I read that normal Devanagari practice (with obviously postdates PaNini)
is to write as a (linearized) lla conjunct, with candrabindu above the
*first* l.
I am not able to get the proper encoding to satisfy this. The nearer I
found is to use la + virama + zwj + candrabindu + la (or the equivalent
ISCII-91 la + halanta + INV + candrabindu + la), but I then have to cheat
when it comes with the short-i matra (because then the code for i have
to come between the zwj/INV and the candrabindu to have correct results).
Anyway, that is certainly a very specific point that does not need
too much or too wide interest.
It is only slightly annoying that we cannot get correct answer.
Antoine
