Re: hex l18n (was: Re: rejection of proposal N2677 (was improper grounds for ...))

From: Philippe Verdy (verdy_p@wanadoo.fr)
Date: Fri Oct 28 2005 - 20:18:55 CST

  • Next message: Philippe Verdy: "Re: Roman Numerals (was Re: Improper grounds for rejection of proposal N2677)"

    From: "Gregg Reynolds" <unicode@arabink.com>
    > Chris Jacobs wrote:
    >>>At 19:48 27/10/2005, Gregg Reynolds wrote:
    >>>> At high school (in Algeria) we used أ ب جـ د هـ و
    >>>>for A B C D E F
    >>>>That's alef, beh, jeem, dal, heh, waw - the first six letters of the
    >>>>traditional abjadia
    >>
    >> And what directionality do they have when used as digits?
    >
    > Now Chris, are you baiting me? ;) (If you look in the archives you'll
    > see the issue of number directionality is near and dear to my heart.)

    My opinion is that RTL Arabic letters (which have strong RTL directionality)
    can only be used to write hex numbers, provided that RTL decimal digits are
    used as well (with strong RTL directionality). But are there RTL decimal
    digits? May be yes, using the Arabic digits as well in a directional
    override, but this would not seem natural, even for Arabic readers. So most
    probably these Arabic letters are used as hex digits by forcing them to
    adopt a LTR directionality, like decimal digits, as if they were behaving
    like Latin letters with just Arabic letter forms, and no Arabic ligatures
    (so they are presented in their isolated forms).

    This requires using BiDi override controls for such numbers in Unicode
    texts. In that case, there's no need to define Latin hex digits (and the use
    of BiDi controls must be implemented in any Arabic shaping engine, so this
    is not a problem). Same remarks for Hebrew. Conclusion: we don't need
    additional codes for Latin hex numbers (because they already have the same
    strong LTR directionality as European decimal digits), and both Arabic and
    Hebrew don't need them as they use BiDi control overrides where appropriate,
    whever they use LTR European decimal digits with their RTL letters, or local
    decimal digits with their RTL letters.

    What Gregg does not show, is how numeric strings *with more than one* digits
    are written in Arabic texts:
    1) decimal or hexadecimal numbers using European digits or Latin letters (I
    thnk there's no problem as this case is identical to European texts)
    2) decimal only numbers using Arabic digits (the traditional form of noting
    numbers in Arabic texts)
    3) hexadecimal numbers using only Arabic letters (are they written RTL or
    LTR? this is the important issue)
    4) hexadecimal numbers using Arabic letters and Arabic decimal digits (most
    probable use of Arabic letters as digits)
    5) very hypothetic case of hexadecimal numbers using European digits and
    Arabic letters (seems strange and not natural, why restrainging on
    converting only letters to Arabic, and keeping European digits?) I think you
    did not learn this as school.
    6) hypothetic (scholar) case of hexadecimal numbers using Arabic digits but
    still European letters (seems strange and not natural, why restrainging on
    converting only digits to Arabic, and keeping European letters?)



    This archive was generated by hypermail 2.1.5 : Fri Oct 28 2005 - 20:21:56 CST