RE: [Indic-computing-devel] Testing OpenType support

[Andy W. <And...@bt...>]

 Keyur Shroff wrote:


> 
> Approach-1:
> ----------
> 
> A1. <C1><virama> will produce Chillu form (if any).
> A2. <C1><virama><ZWNJ> will produce halant form of C1.
> A3. <C1><virama><C2> will produce a ligature if <C1> doesn't
> have chillu form. 
> A4. <C1><virama><ZWJ><C2> will produce a 
> ligature if <C1> has chillu form. 

The problem with A4 here is that the current specification of 'half'
does not allow a ligatures of C1C2. IMHO a new feature would need to be
specified. (feature 'cillu' ?).

> A4 has been given to answer the question:
> "How to produce <C1-chillu><C2> sequence if <C1><virama><C2>
> produce a ligature". As a result the rule A3 should be used 
> when there is no chillu form of <C1> and A4 should be used 
> when there is a chillu form of the consonant <C1>.
> 
> The issue with this approach is that we have to use ZWJ in
> conjunct formation when <C1> has chillu form.

Why not just always put a ZWJ whenever a chillu is desired?
So C1chillu = C1+ZWJ  (always) 
> 
> Approach-2:
> ----------
> 
> B1. <C1><virama><ZWJ> will produce Chillu form (if any).
> B2. <C1><virama> will produce halant form if <C1> is not part
> of a ligature. 
> B3. <C1><virama><ZWNJ> will also produce 
> halant form of <C1>. 
> B4. <C1><virama><C2> will produce a 
> ligature (or consonant conjunct).
> 
> The issue here is that people generally believe that ZWJ is
> used to join the consonant preceding it with the consonant 
> immediately following ZWJ. Thus, ZWJ virtually becomes part 
> of the syllable. Here we are using it to terminate the syllable.

Keyur, I agree with you. Both of the approaches above make sense.
However, Approach-2 is more compatible with the current state of the
situation. It would need less work to get it to work and is less likely
to break any existing implementations (IMHO).
Approach- 2 does not explain how to create a chillu+consonant conjunct.
Assuming your answer is <C1><virama><ZWJ><C2>, then again, either a new
feature definition is required, or the existing feature 'half' will need
respecifying.

Andy