Re: Character Identity and Font Selection

From: Asmus Freytag (
Date: Thu Jun 09 2011 - 00:55:58 CDT

  • Next message: AndrĂ© Szabolcs Szelp: "Re: Character Identity and Font Selection"

    On 6/8/2011 9:28 PM, Michael Everson wrote:
    > On 9 Jun 2011, at 06:21, wrote:
    >> Can anyone remind me why Variation sequences aren't the appropriate response to the IPA problem?
    > That's just pseudo-encoding.

    No, it's not. It depends on how you define the "IPA-problem".

    If you define it as making the distinctions *within* IPA, then, yes, you
    do want these as characters.

    I understood the suggestion to use variation sequences to allow glyph
    selection for those characters that are unified with regular scripts,
    but where IPA restricts the allowable range of glyphs.

    That is glyph selection and for that variation sequences are a very
    proper tool.

    By using variation sequences with these characters you would clearly
    indicate that mapping them to a non-IPA font would violate a glyph
    selection assumption the user had made by using the character in the
    context of IPA.

    Instead of relying on external markup of a passage as IPA, the variation
    selection would allow internal designation of the restricted glyph
    range, something that might be less likely to be separated from the data.

    As such, the idea to complement the existing IPA encoding with such
    variation sequences definitely has merit.

    >> It seems to me that 'a' + VS1 should be defined as the two-story IPA a, ... It would restrict the set of glyphs that should be used to represent the character, but not create a duplicate encoding - exactly what the variation selectors were encoded for, right?
    > Then you have a different glyph but the same character.

    Right. For IPA you only want certain glyphs for 0061, not the full range
    that's permissible in ordinary text.

    That's a restricted glyph range, and precisely the kind of thing for
    which the variation sequences were designed in the first place.

    The reason Unicode doesn't already have these sequences defined is that
    IPA encoding (with all its overloading of 0061 etc) predates the
    recognition that such problems can be addressed in the character stream
    and the subsequent invention of the variation selectors.

    A pure accident of historical development, and not one that must prevent
    the application of variation selectors in a systematic fashion for *all*
    intentional unification of characters that result in such restricted
    glyphic range for some, but not all uses.


    This archive was generated by hypermail 2.1.5 : Thu Jun 09 2011 - 00:59:31 CDT