L2/04-060

From: Gautam Sengupta
Date: 2004-02-01 10:48:22 -0800
Subject: Encoding Bangla Khanda-Ta With Ta+Virama

I would like to submit the following proposal for consideration by the  
UTC. Would any of you know how to send it across? I would also welcome  
comments from you.
Thanks for your help.
-Gautam

============================

PROPOSAL FOR ENCODING THE BANGLA KHANDA-TA WITH TA+VIRAMA

GRAPHOLOGICAL BACKGROUND
In Bangla (=Bengali) a dead consonant TA shows up as KHANDA-TA in all  
contexts except where it is immediately followed by one of the following  
consonants: TA, THA, NA, BA, MA, YA, RA. It is thus NOT a distinct abstract  
character, but a completely predictable display variant of the dead  
consonant TA.
KHANDA-TA cannot bear a vowel matra or combine with a following consonant  
to form a conjunct aksara. It can form a conjunct aksara only with a  
preceding dead consonant RA, with the latter showing up as a REPH placed on  
the KHANDA-TA.

RENDERING
(1) A TA+VIRAMA sequence should always be displayed as a KHANDA-TA except  
when immediately followed by a ZWNJ or one of the consonants listed above.
(2) Any rendering engine should, by default, introduce an orthographic  
syllable break after a TA+VIRAMA sequence except when the pair is  
immediately followed by one of the consonants listed above, or explicitly  
forbidden to do so by an immediately following ZWJ.
This would take care of all normal, orthographically attested occurences  
of KHANDA-TA.

ORTHOGRAPHICALLY UNATTESTED SEQUENCES
To encode a KHANDA-TA+C conjunct or a KHANDA-TA bearing a vowel matra  
(both unattested in normal orthography) a ZWJ should be inserted after the  
TA+VIRAMA sequence, e.g. TA+VIRAMA+ZWJ+RA to encode a KHANDA-TA with a  
subjoined RA (RA-PHOLA) and TA+VIRAMA+ZWJ+MATRA to encode a KHANDA-TA  
bearing a vowel matra.
To force a KHANDA-TA to show up in a context where a TA+VIRAMA would  
normally ligate with a following consonant, a ZWJ followed by a ZWNJ should  
be inserted after the TA+VIRAMA, e.g. TA+VIRAMA+ZWJ+ZWNJ+RA for KHANDA-TA  
followed by base form of RA. NOTE THAT SUCH COMPLICATED SEQUENCES ARE  
NEEDED FOR ENCODING UNATTESTED FORMS ONLY.

MERITS OF THE PROPOSAL
(a) Consistent with the avowed Unicode policy of encoding only abstract  
characters and not display variants.
(b) Only TA+VIRAMA to be used consistently to encode all orthographically  
attested occurences of KHANDA-TA.
(c) ZWJ and ZWNJ required only to encode orthographically unattested  
occurences of "junk sequences".
(d) ZWJ and ZWNJ used in a manner consistent with their conventional semantics.