Re: UTCへのレターの本体

Hi Nat,

Thank you.

> If they could be deemed compatible after adding these 4 characters’ AJ1 postures to UAX50, then this would be fairly easy, but there is another issue: What to do about the different engine logic governing the many U+2xxx and Cyrillic and Greek ranges that originated from SJIS, and their being R in UAX50 and U in AJ1 or SJIS-born fonts.

Here’s what I think.

InDesign
InDesign would continue using its existing logic, which is proven to work for Japanese publishing. As publishing market has a heavy legacy and inertia with its history, I understand it is important to keep every detail of how it has been done on print. With these 4 characters' change, AJ1 fonts do not need to change to adopt UAX50 any more, i.e. it is future compatible and it is a good thing for InDesign.

EPUB
Consequently, many EPUB books would require similar layout. Because we know exactly which characters are affected, we can automatically attach text orientation css to each of such characters. Alternatively, we could propose a globally applicable css text style, e.g. sjis, if we see issues with attaching css upright to every such characters.

Formal definition of so called SJIS style layout
Even if we do not propose such a css style, developing a formal definition of so called SJIS style might be a worthwhile exercise. We could go further and propose an SJIS flavour of UAX50 to Unicode while keeping the original UAX50 for all fonts. Once we’ve done that, we can standardize Japanese text layout into two flavous, UAX50 and SJIS. No random text orientation any more.

Other digital text
For other fields, I honestly do not know how important to maintain the SJIS style, esp. fullwidth treatments for Greek/Cyrillic and math operators, with exception of the first few small letters of Greek characters. Many math operators are new on JIS X 0213 and I have never seen them appearing in Japanese text regardless of the text orientation. Does anyone have knowledge or statistical data on the usage of these characters in Japanese text context?

I guess the specter of the SJIS layout is living only in legacy publishing market. And other fields can migrate to UAX50. That’s my guess but I do not know, and nobody would know. We’ll see. If we find that people expect SJIS style layout even on digital native text, no fear is necessary. We defined what the SJIS style is, and text engines can apply that alternative rule.

幽霊の正体見たり枯れ尾花

- kida

> 2021/08/22 3:02、Nat McCully <nmccully@adobe.com>のメール:
> 
> I think perhaps I was misunderstood about UAX50 and fonts supporting it – I keep thinking of UAX50 as a text layout engine tool, not a font compatibility tool. They are, of course, related, but UAX50 (it seems to me) was not developed with fonts in mind per se, it was developed so text layout engines could decide on a consistent vertical posture for all characters, and using Tr or Tu call on fonts to make appropriate substitutions for any upright characters needing special treatment.
>  
> The issue with fonts is unfortunately that Japanese fonts using a consistent glyph mapping that assumes all engines use UAX50 to determine Tu or Tr transformations, and moreover, that all R glyphs are appropriate for rotated vertical layout, are only now appearing (Source Han/Noto being one example). The only other consistent glyph set and de facto professional standard is AJ1, and fonts using the AJ1 glyph set are (I believe) number in the minority (although they are so only due to cost and level of quality). The highest cost and highest quality fonts are AJ1, but AJ1 is not the only glyph set widely used in the wild in digital media.
>  
> Therefore to ensure that text layout engines (especially those in professional-grade and print industry products) are able to adopt UAX50 more widely, the fonts have to be made compatible with it. If they could be deemed compatible after adding these 4 characters’ AJ1 postures to UAX50, then this would be fairly easy, but there is another issue: What to do about the different engine logic governing the many U+2xxx and Cyrillic and Greek ranges that originated from SJIS, and their being R in UAX50 and U in AJ1 or SJIS-born fonts. We still have the specter of SJIS style layout of unified codepoints made more difficult using UAX50 in the engine logic.
>  
> I agree with a gradual approach, teasing apart the issues one by one, starting with bringing AJ1 and UAX50 closer together by amending UAX50. However I am wondering about the next steps and how achievable those are. Taro and I will try to dig into this further.
>  
> --Nat
>

Received on Saturday, 21 August 2021 23:30:29 UTC