> From: suzmccarth [mailto:suzmccarth@...]


> I read it but I missed the part that might explain that the virama
> has a different nominal glyph (that thing that gets stuck on the
> keyboard) for Devanangari and Tamil. It also has two different
> actions in Devanagari and Tamil, creating conjunct consonants in the
> one and consonant clusters in the other. This information must be
> in Uniscribe since it is not in the encoding.

The virama has a function that, if abstracted, can be generalized across
all the scripts of South Asia (and beyond, in the case of some other
Indic scripts): when it occurs between two consonants, it causes some
conjoining effect. There are several conjoining behaviours, and the
scripts vary according to which of the various behaviours they employ. I
like to characterize them in terms of whether the preceding or following
consonant is altered.

So, assuming an encoded character sequence C1 + virama + C2, the way
most consonant clusters are written can be characterized in terms of the
categories:

C1-conjoining
- half forms (Devanagari and Gujarati)
- reph (several scripts)
C2-conjoining
- vattu (below) forms (typical of Kannada, Telugu, Oriya)
- post forms (e.g. Oriya yya-phalaa)
conjunct ligatures
non-conjoining
- overt halant
- chillu forms

Note that this is a rough typology, and there are plenty of things that
don't fit exactly. For instance, in some languages, reph can be placed
on a following independent vowel, ya-phalaa can be placed on a preceding
independent vowel, or consonants can ligate with a vowel matra* -- the
typology above doesn't attempt to capture all of these.

*BTW, note that terminology for Indic scripts is not used consistently.
The term "matra" is a good example of this. To a Hindi speaker, a matra
is a vowel mark appended to a consonant; to a Bengali speaker, the matra
is the head -- the line across the top.



Peter Constable