xml:lang oddities

[jcsalomon]

The book I’m editing has some foreign-language words, and (mostly so I wouldn’t “fix” these in spell-checking) I applied the XML language tag to them:

<i xml:lang="hu">Én nem beszélek magyarul.</i>

worked ok, but I tried this with some transliterated Korean:

<i xml:lang="ko-Latn">dongsaeng</i>

and that behaved oddly: The desktop version of ADE (version 2) showed the transliterated Korean in an upright font; and Mantano/Bookari displayed them in a slanted roman, rather than an italic, font.

I’ve reported the bug to Mantano, but I’m removing this attempt at semantic coding anyway, since ADE-based readers are just about guaranteed to to something wrong with this. I figured I’d post my results here so anyone trying to research this matter will find this and know not to bother.

(See What is the correct way to use the lang attribute with phonetic pronunciations? on Stack Overflow for an indication that—theoretically, at least—what I tried should have been the Right Thing to Do.™)

[Tex2002ans]

Originally Posted by jcsalomon

[...] but I’m removing this attempt at semantic coding anyway, since ADE-based readers are just about guaranteed to to something wrong with this. I figured I’d post my results here so anyone trying to research this matter will find this and know not to bother.

Do you have any more example code + screenshots of this in action? Maybe attach a sample EPUB showing off the bugs?

I am going to take a giant stab in the dark and suspect this type of bug might only occur for languages that handle emphasis with punctuation marks instead of italic text:

https://en.wikipedia.org/wiki/Emphas...ctuation_marks

Again, another stab in the dark... I assume most of these devices would handle the primary tags ("en", "fr", [...]), and MAYBE the major variants ("en-US", "en-UK", [...]), but not properly handle more obscure variants (like "ko-Latn"), especially where the non-Latin primary language overlaps with some sort of Latin variant (such as you marking up the transliterations).

I bet ADE just sees the "ko" primary language, and automatically treats the entire thing as Korean (because it doesn't understand/distinguish "-Latn").

On a related note, do you suspect this type of bug might also effect some RTL or vertical languages/variants? I don't work with any sorts of those texts to know.