Re: Re[2]: Berber/Tifinagh (was: Swahili & Banthu)

From: Philippe Verdy (
Date: Mon Nov 10 2003 - 16:59:40 EST

  • Next message: Don Osborn: "Re: Berber/Tifinagh (was: Swahili & Banthu)"

    From: "Michael Everson" <>
    > At 17:36 +0300 2003-11-10, Alexander Savenkov wrote:
    > >"The Wrong Thing To Do" can be seen everywhere in the newspapers when
    > >the names and some other words originally written in Cyrillic and
    > >other scripts are letter-by-letter (mapped?) transliterated to the
    > >resulting script.
    > That's transliteration, which is different.

    What other solution will you find for standardizing the Tifinagh script,
    when it clearly appears that glyphs are not relevant for the character

    (1) May be a system similar to ISCII where characters are coded by their
    function rather than their glyph? This is what has been suggested in the
    Tifinagh encoding proposal. It has the caveat that several sets of glyphs
    are needed to represent the same codepoints according to cultural
    conventions, and the additional problem that no convention actually has
    glyphs that cover the whole abstract character set. The code coverage
    however is complete with character names, to which an unambiguous
    accentuated Latin letter is immediately accessible. This is in that model
    that Latin letters would be used as possible glyphs for the script.

    (2) Or a codepoint assignment for each representative glyph? The problem
    would be to name the code points unambiguously (and thus: which cultural
    convention will this name adopt?) This model corresponds more or less to the
    solution used when creating fonts, as a collection of symbols, independantly
    of their actual meaning. The problem is that it becomes difficult to
    interchange texts encoded with each script variant. But if this correspond
    to the script usage for a particular language, and no language uses distinct
    variants simultaneously, this could be a good solution. However the
    character properties will be difficult to define (do they all encode letters
    or syllables?)...

    (3) Or a code point assignment for each (name,glyph) pair ? This solution
    would create multiple code points with the same representative glyph but
    distinct names, and multiple code points with the same basic name or
    function but distinct glyphs. For this case, the chosen Unicode name should
    be qualified by the name of its cultural convention (which one? the name of
    the author who published that convention?)

    (4) Or a set of separate scripts? This would create separate blocks for each
    Tifinagh variant, deunified but with a clear assignment between the glyph
    and its meaning, but would require more codepoints. Here it becomes possible
    to identify the language and script and work on its semantics, with precise
    character properties. This model would correspond to the solution adopted
    when mapping the scripts previously unified in ISCII into Unicode with
    separate code points in separate blocks. Here also the name to use for each
    script block is not clear...

    Fondamentally, the solutions (3) and (4) are nearly equivalent, but solution
    (3) is more compact and mixes the same number of code points into a unique
    block, which would have less "holes".

    This archive was generated by hypermail 2.1.5 : Mon Nov 10 2003 - 17:45:36 EST