Re: New Corrigendum to The Unicode Standard

From: Asmus Freytag (
Date: Fri Aug 17 2007 - 10:44:59 CDT

  • Next message: Omi Azad: "Re: [indic] Re: Feedback on PR-104"

    On 8/17/2007 8:20 AM, Simon Montagu wrote:
    > Asmus Freytag wrote:
    >> On 8/17/2007 6:48 AM, Michael Everson wrote:
    >>> I have to say, I find the behaviour of mirroring punctuation to be
    >>> generally irritating. Even in a simple programme for DVD titles the
    >>> title and transliteration of some Persian behaves unexpectedly.
    >> That's an implementation bug - as long as mirroring is consistently
    >> applied between authoring environment and the time you are viewing
    >> the text, then you should not be able to notice it, even. It should
    >> just work.
    > For some values of "just work", yes, but it's true that the behaviour
    > of mirroring characters at the boundaries of RTL and LTR is often
    > counterintuitive. Even if it can be fixed by LRM/RLM, a typical user
    > entering data doesn't know that, and doesn't know how to enter LRM and
    > RLM anyway.
    Well, I don't write texts in bidi languages, but I've seen that some
    platforms even provide context menus on edit fields so you can enter
    these special characters with a mouse click.
    I know some special characters are even on keyboard layouts, but can't
    tell you offhand which and where.

    But, the problem of counterintuitive behavior at directional boundaries
    is not limited to quotation marks or brackets. It potentially affects
    all punctuation, and I can't imagine that the 'typical' user in those
    language environments doesn't ever pick up on how to enter such texts.

    In a rich text environment, the markup allows even better control of
    directionality - but if the mirroring behavior of common characters
    *changes* then even rich text documents will break - and that's what
    happened in 5.0, and that's why the Corrigendum.

    The lesson learned: whatever warts the bidi solution in Unicode may
    have, radical (or even noticeable) changes are out. (And, in any case,
    any implicit algorithm has trade offs by necessity, so there can be no
    universal optimum anyway).


    This archive was generated by hypermail 2.1.5 : Fri Aug 17 2007 - 10:46:50 CDT