- From: John Daggett <jdaggett@mozilla.com>
- Date: Sun, 15 Jan 2012 20:13:36 -0800 (PST)
- To: koba@antenna.co.jp
- Cc: www-style@w3.org, MURATA Makoto <eb2m-mrt@asahi-net.or.jp>
Tokushige Kobayashi wrote: > > 4. Currently, glyph selection is a function of the underlying font > > engine (eg 'vert' gsub table in OpenType), not CSS (or Unicode) > > Yes. The problem is some fonts do not supply such information as > 'vert', but in the future there will appear more intelligent/good > fonts. > > > Is item 4 the problem under discussion in the blog post? > > Partially, yes. > > But my overall intention is different. > > Unicode TR#4 tries to classify every characters into U, S, SB, T. I > supose it is impossible to classify some characters to one class. > For example, latin alphabet is U for 2.3.2.b.1-i fig. 24, but S for > 2.3.2 b.1-ii fig.25. I think you've misunderstood the intent of UTR50, specifically the East Asian Orientation property. It is *not* defining what "proper" orientation is in all situtations, it is merely trying to define a *default* orientation. For use within CSS, the default orientation will affect default rendering and hopefully will be close to what the most common behavior is for Japanese text. Authors will need to override this by explicitly specifying sideways/upright in cases where the default isn't what they want (i.e. wrapping these textruns in spans styled with text-orientation: upright/sideways). Having a consistent default, rather than leaving it up to implementations to define their own heuristics, is the goal of UTR50. User agents, including the Webkit, IE, and the AH Formatter currently make a decision at the codepoint level about this, in some cases using "very heuristic methods" as you note. The goal here is to have a clear and well-defined default value for the orientation, one that doesn't vary across implementations or fonts. > But the issue is how to select a proper glyph shape and glyph > metrics for each character in vertical writing mode. CSS WG rejects > modern font-technology, and trying to make a original mapping table. > This may cause a serious issue in Japanese typesetting, may become a > show stopper of EPUB in Japan. You'll need to be more clear what you mean by "CSS WG rejects modern font-technology". I think you might have misunderstood how the 'upright-right' property value of the 'text-orientation' is intended to work. With this property value, characters by default have an intrinsic orientation but authors can override this if the default is not suitable for a given context. For spans of Latin text, authors also have the choice of using fullwidth Latin codepoints rather than the basic Latin codepoints, since each will default to upright and sideways respectively. This is a legacy way of doing things but I'm sure there's content out there that controls orientation this way. The problems of defining default orientation clearly is another example of why defining standards requires hard work in specifying small details. Without careful thought, implementations define what they think best which results in different implementations defining behavior that is not interoperable. This is the source of the distress that Koji is hearing from Kadokawa. Regards, John Daggett
Received on Monday, 16 January 2012 04:14:24 UTC