Re: AW: Ligatures fi and ffi

From: Hans Aberg (
Date: Thu Jun 02 2005 - 03:59:55 CDT

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

    At 18:08 -0700 2005/06/01, Asmus Freytag wrote:
    >> > The ligatures would not be needed to be added
    >>> at all, as long as there are simple rules for computing which
    >>> ligatures or other renderings to use, because it would be much better
    >>> to let the computer program to compute the correct rendering.
    >>No, that's not correct. The use of optional ligatures cannot be reasonably
    >>decided on algorithmic grounds alone. Whether you use a ligature for "fi"
    >>or for "st" is a stylistic choice. If you say that ligatures would not be
    >>needed at character level at all, you are saying that typographic styling
    >>must be handled elsewhere. While that's surely a possible view, and shared
    >>by many, it is far from self-evident.
    >It's actually worse than that. In some typographic traditions, such as
    >typesetting German in Fraktur, you have the situation that the semantics
    >of the word define whether or not the style (in this case Fraktur) requires
    >a ligature or prohibits it.

    If there is information which Latin (sub-script)
    and a word lookup list, the process of finding
    the right ligature combination in the or other
    typesetting style can still be automated, for
    example, by a having a word lookup directory.
    Unless it is so that the author actually chooses
    different ligatures in order to communicate
    different semantic information to the reader, it
    is still just a rendering information.

    >If the information about ligatures is left to a separate layer, and can
    >be separated from the text, it cannot be reconstituted without human
    >intervention - and potentially not fully faithfully unless the original
    >author can be consulted.

    The question is really where to put it: As a part
    of an encoding of abstract characters, or as a
    part of some other higher level rendering

    >>Drawing lines between orthography and typography is sometimes very
    >Indeed. This is one of the reasons why people who persist in trying
    >to derive a 'pure' implementation of an abstract encoding model will
    >be bound to fail.

    Most likely, any viable solution will focus on
    finding a reasonable compromise. Examples of such
    compromises are already in the text above. For
    example, one can think of having abstract
    characters of which Latin subscript is used in
    Unicode. Then that can easily be used by a
    renderer to lookup which font and which words
    should use what ligatures. The authors then only
    has to indicate which Latin subscript to use. The
    font selection is done by a higher level computer

    >>It can well be argued that in English, the letter combination
    >>"ae" (in words of Greek or Latin origin) can be written as a ligature
    >>with no change in meaning, as a purely stylistic matter. On the other
    >>hand, in some languages, such a "ligature" is definitely a character on
    >>its own.
    >"Style" is a term that's a bit vague. There are several aspects to it.

    The "ae" ligature example is interesting. In
    English, it does communicate some semantic
    information about the word, namely its heritage
    (etymology). One may discuss how much worth it is
    to communicate some such information. Some words
    in English are written with ligatures, but one
    can also be written without it. Similarly, in
    other subscripts, the choice between "ae" and ""
    <a with dieresis> can in some cases be
    significant, and other cases insignificant.

       Hans Aberg

    This archive was generated by hypermail 2.1.5 : Thu Jun 02 2005 - 04:00:58 CDT