Re: ZWJ, ZWNJ and VS in Latin and other Greek-derived scripts

From: Asmus Freytag (
Date: Thu Jan 25 2007 - 19:18:01 CST

  • Next message: Philippe Verdy: "Re: U+00BA and U+00AA (was: "Re: Public Review Issue Unicode Technical Report #25, "Unicode Support for Mathematics"")"

    On 1/25/2007 4:14 PM, John H. Jenkins wrote:
    > On Jan 25, 2007, at 4:32 PM, Ruszlan Gaszanov wrote:
    >>> (And I should also point out that using ZWJ and ZWNJ to control
    >>> ligation gives
    >>> typographers the heebie-jeebies.)
    >> Surely, decomposed ligatures don't present more problems for
    >> typographers then any
    >> NFD sequence (not to mention Indic and Arabic scripts).
    > The problem is the implication that ligature control belongs in plain
    > text. By and large, for Latin it doesn't. For Latin, it doesn't make
    > much sense to specify ligature formation involving certain characters
    > in the absence of information on what font is being used. Some fonts,
    > like Courier, will typically have no or very few ligatures. Other
    > fonts, such as Zapfino, will have huge ligature repertoires.
    There is that old chestnut again.

    While attempts at arbitrary ligature *formation* via ZWJ has limited
    usefulness in plain text, because it makes assumptions about the fonts
    being used, ligature *prevention* by ZWNJ is a different animal. There
    are several languages where the model "give me all the ligatures in the
    font" results in orthographically incorrect text. In those languages,
    you need a mechanism to prevent ligatures for character pairs in
    specific contexts that are based on the meaning of the word, not its

    As of now, ZWNJ is the only way to do that (and thankfully, no special
    support is needed other than not rendering it with a visible glyph of
    its own).

    The sad reality is that because software cannot automatically ligate
    these languages without errors, the use of ligatures has gone way down.

    > ========
    > John H. Jenkins

    This archive was generated by hypermail 2.1.5 : Thu Jan 25 2007 - 19:19:38 CST