[whatwg] Ruby markup - Furigana

Henri Sivonen hsivonen at iki.fi
Fri Jan 5 02:05:30 PST 2007

On Jan 4, 2007, at 14:51, Michael(tm) Smith wrote:

> Henri Sivonen <hsivonen at iki.fi>, 2007-01-04 14:38 +0200:
>> On Jan 4, 2007, at 12:05, Karl Dubost wrote:
>>> Or read the kanjis that are too difficult to be known when browsing.
>> How does furigana map to aural rendering? Is only the annotation read
>> out loud and the base ignored?
> If by "base" you mean the kanji,

Yes. I meant the normal-sized text on the usual baseline.

> So one of the common uses of furigana (outside of just
> being used in texts for learners who haven't mastered reading yet)
> it to show readings for kanji combinations that are otherwise
> ambiguous. Or to show that a kanji combination should be read
> differently from the way it would otherwise normally be read.

That was my understanding. (Though one has to wonder why they don't  
just use straight hiragana without kanji in cases where the usability  
properties of kanji are so bad that furigana is needed.) So when  
considering what would make sense as a default aural rendering, I  
thought suppressing the ruby base and only reading the ruby text  
might make sense.

http://www.w3.org/TR/ruby/#non-visual is basically a long-winded way  
of saying that the spec writers couldn't come up with a default aural  
rendering and, hence, ruby markup is not media-independent as  
specified. The real problem seems to be that ruby is used both for  
supplementary text and for alternative text.

If it was as simple as this, there should be a flag for indicating  
whether the ruby text is alternative text (in which case aural  
rendering would read it and suppress the base) or whether the ruby  
text is supplementary text (in which case both the base and the ruby  
text should be read with some indication about the ruby text being an  

Next, the problem is what should be the default. The default aural  
rendering should make sense in the common case for documents written  
by visually-oriented authors without requiring the authors to jump  
through semantic hoops, which most of them won't do anyway in  
practice. The W3C spec suggests flagging the alternative text case  
with class='reading', but the spec is woefully inadequate as it  
doesn't normatively specify class='reading'. This make sense, because  
suppressing the aural rendering of the ruby base is potentially  
dataloss. However, if giving the reading is the most common use case,  
the default aural rendering should do the right thing for that use  
case without a special flag and the case where the base needs to be  
read should be the case that requires the author (or WYSIWYG editor!)  
to use a special flag. (Moreover, the W3C samples places the class  
attribute on the ruby text element instead of the wrapper ruby  
element. This isn't nice considering CSS selectors for suppressing  
the ruby base.)

For now, to provoke comments supporting or refuting my totally  
uninformed hypothesis, I am assuming that example of the reading not  
really being the reading (は vs. わ) as explained at http:// 
www.w3.org/TR/ruby/#non-visual is not a real problem (see below). Or  
at least that leaving the problem unsolved and the default aural  
rendering slightly wrong in some cases is better than saying that  
ruby cannot have a reasonable default aural rendering at all.

Here's what makes me suspect that alleged i18n problems may not be  
real problems:

If you ask a random Finn on the street, what letters in addition to  
a–z are needed for writing Finnish, chances are that he says ä and  
ö, but the alphabet also includes å to cover Swedish as well. Now,  
if you ask the people who represent Finland on international  
standards bodies, they'll say you also need š and ž. If you tell  
this to the random Finn from the street, he can come up with use  
cases for š (although he doesn't know how to use the letter on a  
computer and types sh instead) but he'll think claiming that Finnish  
needs ž is just crazy. The sad part is that foreigners have no way of  
knowing that š and ž are an esoteric thing the language people came  
up with and that is of no concern to real end users. So then Red Hat  
et al. actually believe that the sky falls unless they ship an  
ISO-8859-15 Finnish locale instead of an ISO-8859-1 locale (or  
upgrade straight to UTF-8). Trouble ensues.

After finding out what kind of things the Finnish language  
representatives keep telling foreigners as Finnish “requirements”,  
I can only assume that representatives from elsewhere may be giving  
equally detached from reality “requirements” for other languages.  
(And the mostly verifiable rumors I’ve heard about the committee  
behavior of Danish and Dutch representatives support my hypothesis of  
crazy things happening in international standards bodies over alleged  
national requirements.) Since in the Finnish case the expert opinion  
is detached from reality (and I know what the reality with Finnish is  
with utmost confidence), I now suspect expert statements about the  
requirements of other languages as well or I suspect the seriousness  
of alleged problems. :-(

Henri Sivonen
hsivonen at iki.fi

More information about the whatwg mailing list