Re: CSS3 Text Layout: Writing Latin bottom-to-top within line with blocks progressing to the right

On Jan 12, 2009, at 20:28, fantasai wrote:

> Henri Sivonen wrote:
>> On Jan 12, 2009, at 01:03, Alex Mogilevsky wrote:
>>> It's been proposed and was in earlier versions of CSS3 text. It  
>>> has lower priority because there is no live language that would  
>>> use a writing mode like that for normal writing.
>> I see. Without knowing the implementation details, I naïvely think  
>> that supporting the case of rotating Latin text layout 90 counter- 
>> clockwise would have a small incremental implementation cost once  
>> block-progression: rl; is already sunk cost.
>
> It gets complicated once you try to consider what happens to vertical
> scripts in such a layout mode. They can't very well be written
> upside-down. See Unicode Technical Note #22 for some discussion on
> the topic.
>
> Note that there's a difference between vertical text and rotated text.
> In some cases they look the same (pure Latin). In others, they don't
> (Chinese).


It's not cool if vertical column headers with counter-clockwise  
rotated aren't supported in the scripts for which it makes sense just  
because it doesn't make sense for other scripts. One *could* define  
that in this mode, CJK doesn't turn upside down but rotates 90 degrees  
counter-clockwise like non-CJK text just to have *some* well-defined  
behavior that isn't totally crazy.

As a more elaborate solution, CJK could remain top-down without glyph  
rotation leading to a bidi situation (as shown on page 6 of TN22), but  
it doesn't make sense that this bidi complication blocks the  
availability of the unidirectional feature for non-CJK scripts. Also,  
CSS formatters already support bidi for other reasons, so—again naïvely 
—I'm seeing sunk cost and a small incremental feature.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Monday, 12 January 2009 19:15:04 UTC