My name is Glen Vanpelt. From 1984 to 1987 I taught English in Sapporo, on the northern island of Hokkaido, where I learned to read on the subway and to keep my mouth shut at the Snow Festival when people asked whether we had snow in America. Today I translate Japanese technical manuals, mostly printers and machine tools, from a spare bedroom in Portland, Oregon.

Since I put up a small page last year with my name and the word "translator" on it, I've been getting mail. An anime fan in Ohio typed the title of her favorite show in Japanese, and it came out as ‚±‚ñ‚É‚¿‚Í. A host mother in Idaho asked her exchange-student daughter to write a welcome for the family page, and every visitor but the daughter saw a row of little boxes. A high school Japanese club in Georgia got it working on the school computers and nowhere else. They all ask the same thing, a little desperately: what did I do wrong?

Usually nothing, or one small thing. This page is the answer I've been typing into e-mails, all in one place.

—  一  —

Why it turns into garbage

A computer doesn't store letters. It stores numbers, and a list somewhere says which number stands for which letter. For English the list is short, and everyone more or less agrees on it. Japanese uses three scripts at once: hiragana, katakana and kanji, and the standard set of characters runs past six thousand. That needs a much longer list, and each character takes two bytes instead of one.

The trouble is that Japanese computers grew up with more than one list, and they're still all in use:

• Shift_JIS. What Windows and Macintosh computers use in Japan. Most Japanese home pages are written in it.

• EUC-JP. What Unix computers use. Pages made on university and company servers are often in it.

• ISO-2022-JP, often just called "JIS." The old agreement for e-mail, which had to travel through machines that only understood plain seven-bit English. If you send Japanese mail, this is what your mail program should use.

All three hold the same characters. They just number them differently. So when a browser reads a page using the wrong list, or using the English list because nobody told it otherwise, every pair of bytes turns into the wrong thing. The Japanese have a word for the result: 文字化け (mojibake), from moji, "characters," and bakeru, which is what a fox does in the old stories when it disguises itself as something else. Your characters haven't been destroyed. They're in disguise.

You can even tell which disguise. Here is konnichiwa, "hello," the way it comes out when a browser reads it as English:

Low commas and accented letters mean Shift_JIS. A field of ¤ marks means EUC-JP. And if you see a row of narrow katakana and stray punctuation, an EUC-JP page is being read as Shift_JIS. Little empty boxes or question marks are a different problem: the browser knows it's Japanese, but the computer has no Japanese font to draw it with. See what your visitors can do, below.

—  二  —

The one line that fixes most of it

Tell the browser which list you used. You do it with a META tag, and it goes first inside the HEAD, before the TITLE, because the browser needs to know before it reads a single Japanese character, and your title may have some:

<HEAD>
<META HTTP-EQUIV="Content-Type"
 CONTENT="text/html; charset=Shift_JIS">
<TITLE>...</TITLE>
</HEAD>

If your page was saved as EUC-JP, the last word is EUC-JP instead. Spell them exactly that way.

Understand what the tag does and doesn't do. It doesn't change your file. It's a label on the outside of the jar. The file was saved one way or the other by whatever program you typed it in, and the tag has to tell the truth about that. A page saved in EUC-JP and labeled Shift_JIS is a page that lies to the browser in a very official voice, and you'll get garbage that's harder to diagnose than if you'd said nothing at all.

Two more things that cause most of the rest of the trouble:

• Use one list for the whole page. If you paste in a phrase from somewhere else, an e-mail or another page, make sure it went through the same program and came out the same way as the rest.

• Be careful opening a finished Japanese page in an English-only editor. It will show you garbage, which is harmless, but some of them will "fix" the characters when you save, which isn't. Do your Japanese editing in a program that understands Japanese, and keep it there.

—  三  —

What your visitors can do

Even a perfectly labeled page needs a little help from the other end, because most English computers can't show Japanese out of the box. Put a short note on your page, or point people here.

Netscape Navigator 4. Go to the View menu, choose Character Set, then Japanese (Auto-Detect). Navigator will look at the bytes and make a good guess, which helps on pages with no META tag. Navigator still needs a Japanese font on the computer to draw the characters with; once you have one, choose it under Edit, Preferences, Appearance, Fonts, with the encoding set to Japanese.

Internet Explorer 5. The easiest of all. When it meets a Japanese page and the computer can't show it, it offers to download Japanese support for you, fonts and all. Say yes, wait a few minutes, and the page redraws itself. If the guess is ever wrong, the View menu has an Encoding choice that does the same job as Netscape's.

Macintosh. You'll want Apple's Japanese Language Kit, which gives a Mac both the fonts and a way to type.

—  四  —

Typing Japanese on an English Windows computer

You don't need a Japanese keyboard. You need an input program, which the Japanese call an IME, for "input method editor." Microsoft has a free one for English Windows called the Global IME, and there are shareware Japanese word processors such as NJStar. They all work the same way, and once it clicks you'll type faster than you expect:

1. Type the word in our alphabet: nihongo.

2. The program turns it into hiragana as you go: にほんご

3. Press the space bar, and it offers kanji: 日本語. If there's more than one way to write the word, press the space bar again to see the others.

4. Press Enter to accept it.

The skill isn't the typing. It's knowing which kanji is the right one when the program offers you five, and that only comes with reading.

—  五  —

An old courtesy

Whenever you put Japanese on an English page, give the rōmaji beside it, and the meaning: ようこそ (yōkoso, "welcome"). It costs you a few words. Half your visitors will never see the characters, whatever you do, and they'll still know what you meant. The other half are learning, and they'll thank you. When I was new in Sapporo, the kindest thing anybody did for me was a woman at the ward office who wrote the readings in tiny letters beside every kanji on my forms.

And if you only need a phrase or two, think about whether you need the characters at all. Ken Okamoto's Japanese for the Business Traveler gives all fifty of his phrases in rōmaji alone, which neatly sidesteps every problem on this page, and it's exactly right for a traveler, who needs to say the words, not read them.

—  六  —

Twenty words for the Japanese Web

Once your page can show Japanese, you'll want to visit some Japanese home pages, and they have their own vocabulary. I've put twenty of the most useful words, with kana, rōmaji and a note on each, on a page of their own: the glossary.

Visitors since February 1999:

Questions, and garbage you can't decode: [email protected]. Tell me which browser you're using, and paste a line of the garbage into your letter. I can usually tell you what went wrong from that alone.

Last updated September 18, 1999.
© 1999 Glen Vanpelt