Re: DIY OpenType Re-ordering

From: Antoine Leca (
Date: Mon Mar 27 2006 - 03:34:47 CST

  • Next message: Antoine Leca: "Re: Representative glyphs for combining kannada signs"

    Philippe Verdy wrote:
    > From: "Antoine Leca" <>
    >> Philippe Verdy wrote:
    >>> 919, 94D : dead "NG" (NGA + VIRAMA):
    >> More exactly, "dead NGA", or "NGAd". OK.

    >> So you agree this is one, identificated, combing character sequences
    >> (KI being another).
    > I did not say that at that point. This is just what you can say by
    > just looking at those two first characters. The rest needs to be
    > parsed to get the final rendering.

    Not seen yet contextual analysis needed to determine what is a _combing
    character sequences_. Please point to me the text justifying such analysis.

    >>> uses full NGA with subjoined halant by default,
    >> ^^^^^^^^^^
    > Don't protest too soon. This is further discussed.

    You are missing my point here. I just want to insist taht in the present
    Nagari (and more generally Indic) model, the conjoined form (here, stacked)
    is the default, and if one wants to indicate some variation, it is posisble
    to specify looser conjoining, yet I do not see ways to specify "more
    conjoining" behaviour (ZWJ was originally seen as meaning that for other
    scripts, but this was changed early).

    This is a real problem (as it is here) when the more common rendering is
    *not* the more conjoined.

    > Anusvara can't be used to represent a prenasalisation before the
    > FIRST consonnant of a word

    Nor does appear NGA as first letter of any word, much less conjoined with
    KA. Hmmm. You really was thinking it was to be rendered alone in the void?

    > If you use the sanscrit rule, then how would you transcrit in
    > Devanagari the phonetic of the French term "encre"

    You certainly better have to ask a pundit, I am not expert at this at all. I
    do not believe it is easy to match such different phonetics. I would use
    probably अँक्र; perhaps अँक्र्, to represent the French lack of
    prononciation of the final vowel (« e muet »). अंक्र could be correct too,
    and certainly "more Sankrit looking".
    [ɑ̃kʀ] appears quite a good idea to me, too :-).

    >>> If you insert a ZWNJ, it blocks the reordering of I before dead NGA,
    >> How does it links with c.c.s., and "to apply"?
    > I'm not sure about what you mean by "c.c.s.".

    Sorry. "Combing [character] sequences." The idea brung in by Kent, which I
    believed to be out of context here, and to which I thought you were
    commenting about.

    Looks like I was completely outside of context myself.

    > If it is a "combining character sequence" then it is a
    > Unicode term and this is not appropriate here.


    > We are speaking of "grapheme clusters".

    Ah. Thanks for informing me. I guess I should know quit this thread, looks
    like I was displaced away.


    This archive was generated by hypermail 2.1.5 : Mon Mar 27 2006 - 03:36:35 CST