Re: Script of U+0951 .. U+0954

From: Mark Davis (
Date: Mon Dec 09 2002 - 09:48:46 EST

  • Next message: "Re: Possible proposal for new Hebrew accent character"

    In the original design of Scripts.txt, all Mn character were INHERITED, and
    had consistent behavior. However, the UTC wanted to add the script
    information in certain cases, which does complicate the model. (This was not
    affected at all by whether either, both or neither were in ICU.)

    If you wanted to avoid carrying around the general category information, you
    could have a special-purpose table for just the affected Mn characters.

    ► “Eppur si muove” ◄

    ----- Original Message -----
    From: "Doug Ewell" <>
    To: "Unicode Mailing List" <>
    Cc: "Mark Davis" <>
    Sent: Saturday, December 07, 2002 09:15
    Subject: Re: Script of U+0951 .. U+0954

    > There were some errors in my suggested update to Scripts.txt. A
    > correction has been posted. Sorry about that.
    > Mark Davis <mark dot davis at jtcsv dot com> wrote:
    > > Whatever their script property values, characters with general
    > > categories of Mn and Me should also inherit their script from their
    > > base character. The nominal script property value for these characters
    > > may be different from INHERITED in cases where the best interpretation
    > > of that character in isolation would be a specific script.
    > This is more than an explanatory or clarifying passage. It would add
    > noticeable complexity to the Scripts model, because it would now be
    > necessary to distinguish TWO types of "inherited" characters:
    > (1) those marked as belonging to the INHERITED meta-script, which
    > inherit their script from a base character if any, but remain in
    > INHERITED if they occur in isolation (for whatever reason), and
    > (2) those marked as belonging to a "real" script, but with general
    > category Mn or Me, which also inherit their script from a base character
    > if any, but remain in their original script (not INHERITED) if they
    > occur in isolation.
    > Implementations of the Scripts model would need to haul around the
    > general-category information for every character, which was not
    > necessary before and which imposes significant overhead. (Yes, I know
    > ICU already supports this, but suppose I want to roll my own lightweight
    > implementation?) Isn't there some way to keep the "inherited" logic
    > relatively simple?
    > -Doug Ewell
    > Fullerton, California

    This archive was generated by hypermail 2.1.5 : Mon Dec 09 2002 - 12:34:16 EST