4/21/99
Updated 10/3/99

I'm a Japanese minor, and I recently wanted to install Japanese support on my LinuxPPC box so I can type in Japanese. So I checked the web. There aren't that many informative sites, and none had the specific quick-start info I was looking for. Also, it's hard to go very far without running into a Japanese web page, and if you're like me and still learning, those are basically dead ends.

Links:

• The writings of Craig Oda - one of the most informative. Go here for both a general overview and many specific details involved in Japanese computing.
• Excellent description of encodings.
• Jim Breen has an extensive page that has many links to information about computing in Japanese.
• Hiraku Nakajima has a short page about how to use Japanese in the computer abroad.

Input

You need some special way of inputting kana and kanji. There are many choices of protocols, servers, and interfaces, with names like Wnn, umm, canna, kinput2, etc. And then there are different terminal emulators like kterm and rxvt. I'll make it easy for you: get Canna, kinput2, and kterm. To do some Japanese input, all you need is Canna, kinput2, a terminal that can handle kanji, and a Japanese-aware text editor.

If you have a RedHat-based machine (I run LinuxPPC, so that's what the links are to), you can get all of these at ftp://ftp.pc-mind.co.jp or there's another good repository at ftp://core.ring.gr.jp/pub/linux/JRPM/ (with easy-to-find x86 RPMS). Specifically:

• kinput2

  (Make sure you get a version that says "canna" in the file name.) Kinput2 is the thing that starts intercepting your keystrokes when you press shift-space and displays hiragana as you type romanji. When you press the spacebar after typing in some romanji, kinput2 takes what you've typed, asks Canna for options, and then displays the options for you.

  Basic kinput2 instructions:

  • Type shift-space to begin entering kana. Shift-space again terminates kinput2.
  • Begin typing the pronunciation of the kana you want. As you type, it will be converted to hiragana (and it will be underlined).
  • When you are done with a word, press space. At this point, kinput2 makes guesses about what you just entered, converts some of it to kanji, etc.
  • Instead of pressing space, you can press return to instantly accept the currently underlined text.
  • For each highlighted character, choose the kana you want by using the up and down arrow keys. Pressing space a few times will cycle through and then bring up an X window presenting possibilities (you can choose with the mouse if you want). Pressing the left and right keys will move between the kana or words so you can then use the up and down arrows.
  • To extend the selection (of what you're choosing kana for) use ctrl-right (and ctrl-left). Then use the arrow keys or space bar to choose the appropriate kana for the entire selection. It's a little hard to describe, but if you try it you'll see what I mean.
  • Enter little kana by prefixing the pronunciation with an 'x'. For example, chotto is written "cho-xtsu-to" (I added the hyphens for clarity).
  That probably wasn't comprehensive, but it's all I need to know or use. And hey, most documentation for this stuff is in Japanese. I know I would have appreciated this quick guide when I was trying to figure kinput2 out. For more instructions on kinput2 key commands, see Craig Oda's Linux-Nihongo page (link listed above).

• Canna

  Canna is the thing that kinput2 talks to. It's the backend server that takes hiragana and outputs a list of possible kanji. You never interact with it directly; you just need to make sure it's running.

• kterm

  Kterm is an xterm that's built to support the display of kanji fonts. If you're going to be typing into a terminal (for a text editor), you need this or rxvt with kanji support compiled in. Kterm is a bit better than rxvt though (rxvt had font and artifact issues when I tried it).

  Kterm comes set up by default to run kinput2 when you press shift-space. This is good and saves you a lot of headache. Use kterm.

• A text editor

  It's possible to use emacs/mule, but I never got anywhere with it. I use jed, which is damn ugly but works.

• (Recommended) Kanjipad

  This program is neat. You draw a kanji on the canvas (as best you can if you only have a mouse), click the search button, and it gives you a list of the top 5 candidates. You can then click on one and X-paste it into your kterm running jed. For those kanjis you know how to draw but not how to pronounce.

• (Recommended) kakasi

  This program takes stdin and depending on the switches converts it to other formats. The most useful one is kanji->hiragana (or if you're using a non-kterm terminal, kanji->romanji would be very handy).

  Usage: 'kakasi -?' gives you a quick list. What you probably want is 'kakasi -JH -s'. This converts kanji to kana pronunciation (and inserts spaces, which are nice). If you don't have a kterm but want to read Japanese email, for example, use 'kakasi -Ea -Ka -Ha -Ja -s', which converts everything to ASCII.

Trouble? Make sure you're running the servers. It would probably be a good idea to put canna and kinput2 commands in your .Xclients, so they're launched when you start X. You can, of course, launch and kill them normally if you're just experimenting. In order to type in Japanese, you need both kinput2 and canna to be running. You also need to use a multi-byte language aware text editor (like jed-jp) and be using a terminal that 1) can display kana and 2) is set up to start kinput2 when you press shift-space. Kterm comes like this by default.

Printing

Printing was a bitch for me. Here's what I ended up doing:

1. Get VFlib2-2.24.2, ghostscript-5.10-7jp, ghostscript-fonts, nkf (an EUC- to Shift-JIS translator), a2ps (converts Shift-JIS to PostScript). You will need urw-fonts (your standard TrueType font collection) to placate ghostscript's dependencies.
2. Force uninstall ('rpm -e --nodeps') your current ghostscript and ghostscript-fonts.
3. Install the RPMS you just downloaded.

To test it, type a file in your text editor with kinput2. Save it as test.euc (it's EUC encoding). Notice that you can view it in Netscape (as long as you set the Character Encoding correctly).

To convert to PostScript, run nkf to a2ps (ASCII to PostScript) like this: 'nkf -s test.euc | a2ps -nP > test.ps'. Assuming you have Ghostview, you can run 'ghostview test.ps' and see your beautiful kanji.

Important note: So you've got your .ps file and it looks great. Forget about printing it from work. You need access to those kanji ghostscript fonts you installed, and fonts aren't embedded in PostScript files (at least not those made through this procedure). In fact, you might even not be able to print to a PostScript laser printer, since they're supposed to have their own fonts. (There are a bunch of packages in the RPMS directory that contain various extra fonts; you might want to experiment with them.) I use ghostscript to rasterize to my DeskWriter; I've never had any luck with laser printers.