Re: Why people still want to encode precomposed letters

From: Asmus Freytag (
Date: Mon Nov 24 2008 - 13:21:18 CST

  • Next message: Hans Aberg: "Re: Why people still want to encode precomposed letters"

    On 11/23/2008 9:36 PM, John Hudson wrote:
    > Karl Pentzlin wrote:
    >> At least, this. This can be accomplished by an algorithm...
    > An algorithmic approach to mark positioning is certainly a desirable
    > fallback when a font does not contain specific anchor positioning data
    > for the base+mark sequence. This, of course, means that the layout
    > engine will need to query whether the font contains anchor positioning
    > for the glyph sequence -- a blind check for the presence of e.g. a
    > <mark> layout feature in an OpenType font is not sufficient, since such
    > a feature may support e.g. Latin+mark combinations but not Cyrillic+mark
    > --, and if not then apply the algorithm.
    > There are, of course, things that the algorithmic approach won't be able
    > to handle reliably: marks that attach to the base in different
    > horizontal positions depending on the shape of base (e.g. ogonek), marks
    > that are by design not centred but e.g. above the right side of the
    > lettter, and the substitution of flatter accent forms above capital and
    > ascending letters.

    I think that an algorithm could have generic tables that capture some
    aspects of typical ink distribution for certain characters. For example,
    the generic fact that the typical A is trigonal, whereas O or U etc.
    typically touch the base only in the center. Having some canned
    information like that beats having information for each individual pair.
    If this info fails for any specific font - well, it's intended as a
    better fallback.

    I also think that if such algorithms were more widely deployed, it would
    be advantageous if there was a way for font suppliers to create and make
    accessible generic flattened forms of diacritics. That's still cheaper
    than designing all the specific combinations.


    This archive was generated by hypermail 2.1.5 : Mon Nov 24 2008 - 13:23:53 CST