When something is just flipping wrong

by Michael S. Kaplan, published on 2010/03/01 07:01 -05:00, original URI: http://blogs.msdn.com/b/michkap/archive/2010/03/01/9970601.aspx

There are several people who have appreciated the fact that I am cursing less here, lately.

They may applaud my title choice here as it does appear that I have replaced an offensive-to-some Anglo-Saxon monosyllable also known as "the F word" with a replacement.

Alas that is not what the title does, except as a "conceptual homonym" kind of thing. I meant flipping more literally, actually.

It started the other day, when there were some free t-shirts in the hall at the office.

They weren't company branded or anything, they were just something with a cartoon character saying something.

I didn't take one (too many t-shirts already!), but my Japanese colleague took one as she noticed something interesting, as she said in mail to the team later:

Some of us got a brown t-shirt today. The t-shirt has Japanese paint and characters on it. I found a WR issue which happens a lot.
They would need a Japanese font for a vertical writing. Who knows we have Japanese font for both vertical writing and horizontal writing?

I started kicking myself for not grabbing one of the shirts!

Luckily she also took a picture for us to see:

After I expressed my interest she explained the joke here:

The texts show a (poor) joke. The sentence means ”A team which robot “Haiku” visits goes!” The  pronunciation of “Haiku” is the same as “(subject) go(es)” in Japanese, so that sentence contains 2 same pronunciations with different meanings.

This mistake happens a lot. I give you another example. That’s the Mariners Official goods “Ichiro Bear.” It has Ichiro’s name in Japanese on it but used a wrong character for the vertical writing:

A homonym-based joke in the t-shirt, a-la-Semper ubi sub ubi? Cool!

Though the mistake makes it seem less interesting somehow.

So let's take some Japanese text. This text, say:


Now let's look at it vertically (how well this works for you may depend on your browser):


Note how that last Kana, the cho-on, is a vertical line -- it changes in orientation here.

It might help if I make the text smaller here:



There is a very small bend in the end of the cho-on in many Japanese fonts (including most of the ones on my machine), one that disappears at larger sizes. Maybe you can see the difference?

Certainly the text on the bear shows it -- and reveals perhaps a clue as to how it is wrong!

Anyway, one cannot tell exactly what is going on here since we lack access to the presumably computerized typesetting information, but look at these characters for a second.

First Horizontally, than vertically:



Those three are:

Now what is important to note here?

Several things:

  1. The fact that the first one does the vertical thing in a totally different way than the other two.
  2. The different direction of that small bend in the various characters when it is there.

Now you can perhaps do some forensic typography among these and the various other ideographic and kana characters to help determine what is going on (whether it is incomplete vertical support, a font error, or simply the wrong character being inserted since it will look "almost right" or maybe "right enough" in some circumstances).

It is fair to say that vertical support still has a ways to go before people can avoid such investigations to determine what might be behind the problems here, in any case.

Though in both of the examples above, it is most likely vendor's character selection error. This tendency to sometimes grab whatever character might look right without knowing how it will look when vertical support kicks in is one thing that can keep platform support from working 100% properly....

Edson Watanabe on 1 Mar 2010 10:35 AM:

I've tried your page ( http://blogs.msdn.com/michkap/archive/2010/03/01/9970601.aspx )  in Mozilla Firefox, Google Chrome and Apple Safari browsers in Windows XP SP3 and all of them don't render correctly the chōon symbol with vertical text. Only Internet Explorer (sigh) does the correct transformation.

Michael S. Kaplan on 1 Mar 2010 11:26 AM:

That is disappointing news, clearly my concerns were understated, if anything. :-(

jmdesp on 1 Mar 2010 3:10 PM:

Mike, if I'm not mistaken, you are cheating vertical text, by making the width of the div where the text displays so short that each character is on the next line therefore appears under the previous one.

I have no idea how IE is doing to guess your intent and change the orientation of cho-on in that case.

You should be using the CSS3 block-progression (or writing-mode) property to do this properly.

Now, neither Firefox, nor Chrome are handling this property correctly, opposite to IE :-)

There's bug 145503 open about that in Firefox (no real progress currently)

Steve Williams on 1 Mar 2010 3:35 PM:

I was reading the blog in Outlook's RSS reader and both lines of text were identical, i.e. horizontal.  I then viewed it in Firefox 3.6.  It tried to do vertical, but ended up placin the glyphs over each other.  In IE 8, the text was vertical, but the large typeface was clipped on the left side.  The smaller typeface was displayed correctly.

Ahh, the fun. :)

Michael S. Kaplan on 1 Mar 2010 6:24 PM:

Hey jmdesp! I got the hint here and just went with it. Like alchemy! Though I got some insight into how IE does the orientation change -- it is built into code IE uses to read it from the font....

Andrew West on 2 Mar 2010 6:06 AM:

I think that the magic works because the font has two glyphs for U+30FC, one for horizontal layout (a horizontal stroke) and one for vertical layout (a vertical stroke), and when "writing-mode : tb-rl" is applied the correct glyph for vertical layout is selected. So with fonts like MingliU which have only one glyph for U+30FC, the character is rendered as a horizontal stroke in both horizontal and vertical layout.

The same problem occurs with U+3127 (Bopomofo Letter I), although it should be a vertical stroke in horizontal layout and a horizontal stroke in vertical layout, which is the opposite way round to U+30FC. However, I cannot find a single font that renders correctly for both horizontal layout and vertical layout. It seems that the solution used by Microsoft bas been to use a vertical stroke for U+3127 in PRC-targetted fonts (e.g. Microsoft YaHei) which are assumed to be used in horizontal layout only, but to use a horizontal stroke for U+3127 in Taiwan-targetted fonts (e.g. Microsoft JhengHei) which are assumed to be used in vertical layout (at least for typesetting). The result is that the wrong glyph (a horizontal stroke) is normally seen in ordinary horizontal bopomofo text on the internet (e.g. http://en.wikipedia.org/wiki/Bopomofo). This issue was the reason behind Taiwan proposing to encode separate horizontal and vertical forms of Bopomofo Letter I a couple of years ago (http://std.dkuug.dk/jtc1/sc2/wg2/docs/n3246.pdf), but luckily that proposal did not get very far.

Michael S. Kaplan on 2 Mar 2010 6:39 AM:

I have learned many interesting details about innards during the last 24 hours that will possibly find their way into blogs, at some point. :)

Please consider a donation to keep this archive running, maintained and free of advertising.
Donate €20 or more to receive an offline copy of the whole archive including all images.

referenced by

2012/09/07 Let's stand up a moment. You know, Vertically...

2010/09/22 Looking at life a bit more vertically, for a moment...

go to newer or older post, or back to index or month or day