RE: Display of Mongolian in Arabic or Hebrew documents

From: Peter Constable (
Date: Sat Nov 24 2007 - 12:24:16 CST

  • Next message: Peter Constable: "RE: Ol Chiki character name typo?"

    > From: [] On
    > Behalf Of "André Szabolcs Szelp"

    > In the most probable scenario, the user of inline
    > Mongolian words in Arabic main text would be a similar one, i.e. an
    > Arabic/Farsi etc. orientalist.

    Maybe; maybe not, if there is any significant level of contact between Uighur and Mongolian communities.

    > It is my opinion, that it serves that group, if the reading direction
    > is not changed.

    There may be several factors in what serves them best. We're still talking off the tops of our heads by making assumptions.

    I agree that it is a plausible scenario that what may serve some users best would be Mongolian embedded in Arabic in RTL order. I haven't argued against that. What I have said is that this probably going to take some extra work for implementers, that I've yet to see a demonstrated need, that I've seen more demonstrated need for other layouts such as short vertical runs in a horizontal line, and that in most implementations these options are going to involve support in a text stack at a level above run layout.

    > Taking as an example Latin ("when embedding Latin, you also change
    > reading direction") is not a valid argument, as in the case of Latin
    > it's a choice of "no change" vs. "180° change". In the case of
    > Mongolian, it's the choice of "90° change" vs. "90° change" -- CCW or
    > CW.

    An argument for CW rotation of Mongolian when embedding into Arabic was made on the basis of historical derivation of Mongolian script. I made reference to Latin *only* to make the point that this *historical argument* is invalid. It seems I haven't explained that clearly since you keep coming back to this point, which is (more or less) that Latin doesn't provide an argument for what is most natural for Arabic readers. But I was never making such an argument.

    > I personally believe, that every script's glyph should be designed in
    > their original orientation.

    What does it matter? The vast majority of users have no idea whatsoever what all is involved in getting their letterforms to display on the screen, let alone what orientation is used at glyph design time. This is just purism for the sake of purism, expending resources on implementation that provide no benefit whatsoever to end users.

    And if people are really concerned about Mongolian type designers, then the most cost effective solution is to make glyph design tools in which the design surface can be rotated in the UI. That makes far more sense than expecting changes at every level in existing text stack implementations.

    > Mongolian -- as opposed to Chinese -- is
    > always written top-to-bottom natively. Turning the glyphs in embedding
    > should be part of the standard, and not as someone claimed the duty of
    > a higher protocol. You don't use a higher protocol for embedding, say,
    > Armenian into, say, Hebrew.

    While you might not agree with it, Mongolian layout, especially in mixed-script and layout-directionality scenarios, is outside the current scope of the Unicode Standard (if that's what you meant by the "standard"). The Unicode Standard has relatively little to say about text display; the Bidi Algorithm is the only detailed statement, and its scope is limited to how layout should be done when combining text of two horizontal directions. If Mongolian were written, or treated as being written, horizontally LTR or RTL, then the Bidi Algorithm says how it should be combined with text in other horizontal scripts. The Unicode Standard does *not* say how text in a normally-vertical script should be embedded into horizontal text flow, or vice versa -- and probably for a reason: even considering Latin presented in vertical layout, there's more than one way it can be and is done.

    > The current practice of designing Mongolian glyphs as if they were
    > left-to-right oriented is a misbegotten solution, a compromise to the
    > capabilities of current technology.

    It is a low-cost way to build off existing technologies. You can call that a compromise -- I won't debate that. But as I see it, it's the only realistic way to expect widespread support for Mongolian -- comparable to the kinds of decisions that had to be taken early in the history of Unicode to accommodate existing legacy implementations to have any hope of success.

    > Our aim, however, should be to
    > create a standard, which is then adopted by technology, and not to
    > create a standard to fit the current whims of it.

    I am indeed suggesting to create the Unicode Standard and related implementation conventions so that they are adopted. It seems to me you're suggesting that the Standard or implementation conventions be created more for theoretical purism than so that they will be adopted in practice, and to me that is whimsical.

    > The abovementioned current solution seems fair, for short words and for
    > the most common Latin-centric use of the fonts. However, it's already
    > not apt for writing Mongolian: for a native use of the fonts, even RTL
    > oriented glyphs would be better (as long as there are not layout
    > engines to handle vertically designed glyphs in vertical orientation):
    > in an RTL designed Mongolian font the (native) user of it could write a
    > document RTL in his native script (with wrong orientation on the
    > screen), but when printed, the page can be turned, and the book bound
    > the native-orientation-wise.
    > With the current setup of the fonts this cannot be achieved.

    In the short term, designing Mongolian glyphs in RTL orientation would allow a Mongolian user to author content using existing horizontal-layout applications. In the long term, that will make support for vertical editing of Mongolian in general-purpose (i.e., not Mongolian-specific) applications more costly and hence take longer to be supported, if supported at all.

    > It should be the primary goal to create layout engines which can handle
    > TTB Mongolian properly and the glyphs should be designed in their
    > original orientation.

    I agree with the first part: it should a primary goal (in terms of Mongolian support) to create layout engines that can handle TTB Mongolian properly.

    I completely disagree with the second part: requiring that Mongolian glyphs be designed in vertical orientation has no direct connection to users, and so should not be a priority.

    > For embedding, if no definite tradition exists, the embedding should
    > follow the direction of the main script.

    Purism with no connection to demonstrated user need. Priority for how to support embedding Mongolian into horizontal text should be driven by end-to-end user scenarios based on known needs.

    > As a note, one of the scans showed TTB Chinese with embedded Arabic
    > turned CW, so that it would run BTT. That example is (IMHO) irrelevant,
    > because if you look carefully, the very same document contains embedded
    > Latin as well. It is clear, that in the case of embedding horizontal
    > scripts (regardless of them being RTL or LTR) into vertical, the
    > turning would be conducted into the same direction, so that those
    > scripts retain their traditional direction in regard with each other,
    > so that they'd feature their baselines on the same side.

    I agree that that makes sense. But note what happens if you apply that reasoning for embedding Chinese and Mongolian into horizontal text: you either get LTR Chinese and Mongolian, or RTL Chinese and Mongolian. Anybody for Chinese embedded RTL in Arabic (presumably with glyphs rotated 90° CW)?


    This archive was generated by hypermail 2.1.5 : Sat Nov 24 2007 - 12:26:07 CST