[css3-writing-modes] text-combine and half/third/fourth-width glyphs

> The one general comment I would have is that the rendering behavior must be consistent
> across user agents, allowing user agents to pick and choose which behavior to implement
> is a recipe for inconsistency and a huge headache for authors.
> 
> For example, if text-combine is applied to a three character span containing "123" then
> there should be a way of assuring that third-width glyphs are *always* used if an a font
> containing third-width glyphs is provided.  This should work across implementations, it
> should *not* be up to the implementation to decide to wing it and render using shrunken
> glyphs from MS PGothic (double ick!).

I understand your point. This is tough part, and making this behavior UA dependent is actually from requests from implementers.

There can be cases where implementers would like to choose not to use half/third/fourth-width glyphs. For example:

1. The font has proportional digits and thus "11" fits within 1em square box. Since half-width glyphs usually look worse than regular glyphs, regular glyphs should be preferred.

2. The span contains "1i" and its width is within 1em square box. Same as #1.

3. The span contains "1W" and its width exceeds 1em square box drastically. Half-width glyph is available for "1", but not for "W", and even with half-width glyph is used for "1", it still exceeds 1em square box drastically. Combining half-width glyph of "1" and scaled "W" may look better in some cases, while just scaling "1W" may look better. It depends on fonts and rasterizers.

It must be difficult to optimize these cases on PC platforms where browsers cannot assume anything about used fonts, but for devices, fonts and rasterizers are limited and users may not install new fonts. In such cases, implementers would like to optimize their algorithms for these cases for the environment.

While I agree with what you said, the most important feature of text-combine is to fit the combined glyph to almost 1em square box. As long as that part is consistent across browsers, I'd like to allow implementers to optimize their algorithms rather than forcing one rule that is good for one platform but may not be the best for other platforms.


Regards,
Koji

Received on Wednesday, 19 January 2011 20:37:11 UTC