Re: Ligatures fi and ffi

From: Antoine Leca (
Date: Wed Jun 01 2005 - 10:21:30 CDT

  • Next message: Hans Aberg: "Re: Ligatures fi and ffi"

    On Wednesday, June 1st, 2005 12:21Z Dominikus Scherkl wrote:

    >>> - - ligature processing is a required feature to support
    >>> even legacy ISO 8859 charsets like Arabic, or Indian
    >>> standard charsets (ISCII).

    Philippe, please remember that ISCII does not imply a script. In fact, RMN
    (that is, Latin script) is an available choice to render ISCII.

    >> Pardon? In which sense is ligature processing _required_? Do
    >> you mean that it is forbidden now to render "f" followed by
    >> "i" as two letters, without using a ligature?
    > _required_ not for latin, but for Arabic or Indian.

    I cannot comment about Arabic, but "Indian" is too vague a word. Situation
    is quite different when you consider either Devanagari (where there _is_
    required ligatures, although not as many as you might believe) or Tamil
    (where according to a recent message of a native person, the almost only
    known ligature, ksha, is now often rendered with separate glyphs; and anyway
    it is not a genuine character of the script, it is used for "foreign"

    > Those languages become almost unreadable without ligatures,

    And here we skip from scripts to languages, which make the point even

    > and they using hundreds and thousands of them.

    Only the finest typography in Bengali script uses more than 1,000 ligatures,
    and the figure I have in mind is about 1,700, not "thousands." And you can
    certainly achieve readibility with much less glyphs, even if 220 is too low
    a number.
    I also doubt that Arabic uses much more ligatures.

    Unless of course you aggregate all the scripts of India into an unique
    "Indian script." But then I want to see a font that renders this "script" in
    an unique style, that is, _not_ a pot-pourri of various fonts indexed by
    Unicode values, like is Arial Unicode MS. The best candidates for me are
    Code 2000 and Everson Mono, but the first shows some variation in style, and
    is unable to correctly join e.g. U+0915 with U+09BE; and I never actually
    see the second.


    This archive was generated by hypermail 2.1.5 : Wed Jun 01 2005 - 10:22:39 CDT