Re: Fixed Width Spaces (was: Printing and Displaying Dependent Vowels)

From: Asmus Freytag (
Date: Tue Mar 30 2004 - 14:46:41 EST

  • Next message: D. Starner: "Re: What is the principle?"

    At 10:12 AM 3/30/2004, Ernest Cline wrote:

    > > [Original Message]
    > > From: Asmus Freytag <>
    > >
    > > At 12:19 PM 3/29/2004, Ernest Cline wrote:
    > >
    > > >
    > > >UAX #14 makes a rather definitive statement on this issue, albeit
    > > >in an obscure place, in Section 3: Introduction.
    > >
    > > 4.0.1 will amend that section to correct the wrong impression that NBSP is
    > > fixed width and to clarify that this statement is not intended to cover
    > > specialized cases, but just ordinary typographical conventions:
    > >
    > > I'm sorry if the fact that the placement and context of text was not
    > > to guide the reader. Note that the 'obscure place' was in the
    > > introduction (!) of the UAX, where it was a mere note on a subject not
    > > actually covered by the UAX (i.e. line layout) that nevertheless forms
    > > the context in which linebreaking happens.
    >True, but it was the only guidance on the subject that is present in
    >Unicode 4.0.0, and there do exist widely used applications that do
    >treat NBSP as a fixed width space.

    Can you elaborate?

    HTML treats it as a fixed space in terms of the underlying source
    (spaces and line endings, but not NBSP can be coalesced under some
    circumstances in HTML). But that's a source file issue, not a statement
    on how to treat the NBSP in line justification.

    So, if you can, please provide more details as to when and where and how.

    >Still, there is a need for a fixed width space with a width equal to the
    >unjustified width of a normal space .

    Perhaps you would like to elaborate where and when that is used. What
    problem does a fixed width space solve? Are those circumstances where
    it flows with the line, or are those uses limited to tables?

    >With NBSP being ruled out
    >for that job, that leaves FIGURE SPACE, MMSP, and FOUR-PER-EM
    >SPACE as the closest alternatives, but none of them are guaranteed
    >to be exactly that width, even if they are available.

    Figure space should be as wide as a digit, MMSP is something else,
    and I don't know whether FOUR-PER-EM is the width of a typical space.

    >I suppose
    >suspending justification for just one space via a higher-level
    >protocol could work, except I'm not aware of any such protocol that
    >works at a fine-enough grain to do that.

    Justification is usually a paragraph level property and therefore
    if a fixed width space is required, that would be the wrong approach.

    >Also, one could by that argument
    >also argue that many of the current fixed width spaces could be
    >handled by higher level protocol as well.
    >Perhaps a possible U+2064 NONJUSTIFYING SPACE would make
    >sense with line breaking class BA like most of the other fixed width
    >spaces. (I would have preferred proposing U+205E to place it
    >adjoining MMSP, but that code point is already in the pipeline.)

    It might be useful if you could write this (and the supporting evidence
    for it) up as proposal. That would give the basis for either recognizing
    a missing character and adding it, or, alternatively for UTC to recommend
    some other practice.


    This archive was generated by hypermail 2.1.5 : Tue Mar 30 2004 - 15:42:42 EST