RE: [css3-writing-modes] text-orientation: upright (was RE: Minutes and Resolutions 2011-07-13

Thank you Michel.

I agree that EAW and Script are not very reliable. Fantasai and I spent a long time to come up with the current definition[1] using UnicodeSet utility[2] and we even built our own tools to analyze these properties to see how much they match to our needs. If you read Appendix C carefully, I hope you'll notice that we're not blindly relying on Unicode. We tried to use EAW only when we think it's safe to use. When it's really ambiguous, we made them sideways because then authors can use text-orientation: upright to "fix" it.

I think we can summarize current issues regarding the text-orientation to two topics:

1. How much level of features we want for writing-modes level 3. Current ED matches to the original spec, to what most apps provides (except a few high-end apps,) and to what Windows/MacOS provides as an OS feature. But there are opinions we should do more in level 3.

2. The quality of "default good rule" defined in the current ED[1]. Most app/OSes mentioned above relies on heuristic rules such as "name table contains CJK characters" or "if the font has Shift-JIS cmap" to determine glyph orientation. Also, most apps/OSes built the feature before EAW was defined, so the current ED is probably the first try to determine glyph orientation using combination of Unicode properties, instead of relying on such heuristic rules. I'm not surprised if such a first try in the world is questioned.

For the former issue, I think we can discuss further, but my opinion is to stick on the current level. Not only EPUB but also a lot of people in East Asia are waiting for the spec to be done. Making this spec CR/PR/REC earlier is more important than having more features from my perspective.

For the latter issue, this is really a tough one. It must take long time even to review current rules as it took really long time for us to come up with this list, so current fantasai and my opinion is to make it non-normative and wait for implementers to look into this. But if anyone is willing to read through the list and try to figure out any possible issues with the definition, I'd really appreciate.

[1] http://dev.w3.org/csswg/css3-writing-modes/#vertical-typesetting-details
[2] http://unicode.org/cldr/utility/list-unicodeset.jsp?a=%0D%0A&g=

Regards,
Koji

-----Original Message-----
From: Michel Suignard [mailto:michel@suignard.com] 
Sent: Saturday, July 16, 2011 1:53 AM
To: Koji Ishii
Cc: Stephen Zilles; www-style@w3.org
Subject: RE: [css3-writing-modes] text-orientation: upright (was RE: Minutes and Resolutions 2011-07-13

> From Koji Ishii
> (quoting the spec in [1])
>  * If the font and font system support mixed-orientation typesetting
>  (e.g. the Open >Type font used has the vrt2), the UA should rely on
> that feature to set 'vertical-right' text. Similarly if the font and font
>  system support upright typesetting (e.g. the OpenType font used
> has the vert feature) then the UA should rely on that feature to
> set 'upright' text.
>
>
It was for sure the intent of the original writing-mode spec in vertical mode to reproduce in its default value the behavior available on the Windows platform using the OpenType features and the '@font' selection in GDI. It works well in East Asian context with an East Asian font. 
>
>* If the UA needs to synthesize such features, then the settings in Appendix C are recommended.
>
> Appendix C is not easy, probably takes hours to review and understand.
>  I feel sorry about that, but it can't be helped because the feature isn't easy.

No doubt that emulating CJK vertical writing w/o OS and Font support is fiendishly hard. Many things need to be done right and typographic expectations are different between CJK constituencies that make that even harder. Using fonts dedicated to each market makes this easier.

Commenting on Annex C:
- East Asian width is a mere indicator of what should be wide or upright in vertical mode. The value is only an indication for symbols because it is not constant among East Asian fonts either through versions of the same font or across countries. In my last years as implementer I learnt to pretty much ignore the 'Ambiguous' value and override the other default value in many instances.
- The Unicode script value is vastly abused. It was never intended to be used straight as a mechanism to establish typographically correct layout. I don't mean it is not possible but you need to introduce a fairly large amount of analysis above it to make a decent job, especially when dealing with the common and inherit values.
- Same for the Unicode General Category property. 

Overall, I am glad that finally this part of the spec is coming to adoption after all these years. I am not active on this domain anymore but have kept interest in CSS work and I am following this list.

Michel Suignard

[1] http://dev.w3.org/csswg/css3-writing-modes/#text-orientation
[2] http://lists.w3.org/Archives/Public/www-style/2011Jul/0205.html
[3] http://dev.w3.org/csswg/css3-writing-modes/#vertical-typesetting-details

Received on Tuesday, 19 July 2011 09:22:22 UTC