Last gasp of lead type

« previous post | next post »

Chen Cheng-wei and staff writer Elizabeth Hsu, "Taiwan's last lead-character mold maker works to preserve the past" (Focus Taiwan, 5/1/17):

Rixing Type Foundry is home to the last remaining collection of traditional Chinese movable type character molds in the world. It possesses 120,000 molds of different characters in a wide range of sizes and three different typefaces — kaiti (楷體) or regular script, songti (宋體) and heiti (黒體) or sans-serif black — and has more than 10 million lead character pieces for printing or sale.

It is hard enough to design electronic characters:

"Character building is costly and time consuming" (12/22/15)

"How to generate fake Chinese characters automatically" (12/30/15)

"Recurrent Net Dreams Up Fake Chinese Characters in Vector Format with TensorFlow" (otoro, 12/28/15)

"How many more Chinese characters are needed?" (10/25/16)

Making and managing lead fonts consisting of tens of thousands of characters is an incredibly challenging task.  Think 26 multiplied by 1,000 for the number of discrete characters in an encyclopedic dictionary, nearly all of which are far more complicated than each of the letters of the alphabet (cf. a, b, c with ā 錒 ["actinium"], bì 斃 ["kill; die violent death"], xī 巇 ["a crack; hazardous"]).  How do you organize them?  How do you find them?

As painful testimony of the excruciating difficulty of maintaining a lead typesetting font for Chinese, I will give here a very brief account of my own experience with the Harvard Journal of Asian Studies (HJAS).

Sometime before I arrived at Harvard in the early seventies, the Harvard-Yenching Institute acquired a complete Chinese font for typesetting the journal and other publications.  It is very interesting to note that they purchased the font and along with it came a Japanese professional to work full time as typesetter.  This he did for the next several decades.

The H-Y Chinese font occupied an entire room and consisted of an enormous number of individual lead types.  I think the total amount was similar to that of the Rixing fonts mentioned about — around 120,000 individual pieces in different sizes and faces.

When I started publishing reviews and articles at HJAS in the mid-seventies, the journal still looked all right.  But as the years passed, the font got out of order and many of the individual pieces of type could no longer be found.  In the pages of HJAS, those missing pieces of type were substituted by circles (i.e., blanks) and a Morohashi number.

As the eighties wore on, the number of circles in my articles increased to the point that it was annoying and then downright aggravating.  Finally, I told the editor that I would never publish in HJAS again until the situation was rectified.  It was simply too embarrassing to have my readers constantly turn through the pages of the mammoth 13 volumes of the Morohashi dictionary to find which characters were missing.  And the pages of HJAS looked simply ghastly, as though riddled with bullet holes.  Finally, the editor arranged to have the journal electronically typeset in Korea.  After that, HJAS looked much cleaner and neater.  Furthermore, it cost less to have the typesetting done well in Korea than to have it done shoddily in Cambridge, Massachusetts.

Still more Sinological suffering!

"Sinological suffering" (3/31/17)

"More Sinological suffering" (4/3/17)

For a miniature version of the problem, see:

"Chinese Typewriter" (6/30/09)

"Chinese typewriter, part 2" (4/17/11)

[h.t. Mark Swofford]


  1. Jim Breen said,

    May 2, 2017 @ 7:13 pm

    I was gobsmacked to read that the HJAS used Morohashi numbers to index missing hanzi. I know that Morohashi is regarded by some kanji scholars as having a rather Chinese bias, but is/was there no comparable hanzi 字典 that could have been used instead?

    At least the HJAS used hanzi. For far too long many non-Japanese scholarly publications on Japanese matters relied solely on romanization. In early 2000 I was asked to do a paper on my electronic dictionary work for the Japanese Studies Association of Australia's "Japanese Studies". I produced the paper, then the message came back that the printing company couldn't handle kanji and kana, so could I please convert the paper to romaji, I dug my heels in, and made some probably intemperate remarks about the approach of the 21st century, the decades-old availability of Japanese typesetting software, etc., etc. Anyway, it turned out that the printer was able to upgrade their system and the paper duly appeared as written. It probably only took a bit of a push to make it happen,

  2. Carl said,

    May 2, 2017 @ 10:14 pm

    Computerization has been net positive for Asian typesetting but there was a rough period from the 80s to early 2000s in which electronically typeset article have ghastly looking hanzi/kanji. Often there would just be a page of handwritten characters in the back, or worse clearly handwritten characters interspersed throughout! By contrast, books published in Japan in the 30s through 50s have beautiful typesetting with traditional characters and kana usage. So it goes! Perhaps someday they will finally make vertical text layouts work in web browsers.

  3. AntC said,

    May 3, 2017 @ 12:10 am

    Thanks @Victor. I was taken aback by these remarks:

    a complete Chinese font for typesetting the journal and other publications. … purchased the font and along with it came a Japanese professional to work full time as typesetter.

    Finally … the journal electronically typeset in Korea.

    Why a Japanese professional for a Chinese font? Why a Korean typesetter? (Weren't there expat Chinese typesetters in America?)

    I suppose also: why a Japanese numbering system (Morohashi) for Chinese characters?

  4. Bathrobe said,

    May 3, 2017 @ 12:15 am

    Don't vertical text layouts already work in web browsers? This is the CSS I use for Japanese:

    .vertical {
    -moz-writing-mode: vertical-rl;
    writing-mode: vertical-rl;
    -webkit-writing-mode: vertical-rl;
    -o-writing-mode: vertical-rl;
    -ms-writing-mode: tb-rl;
    writing-mode: tb-rl;
    -moz-writing-mode: upright;
    writing-mode: upright;
    -webkit-writing-mode: upright;
    -o-writing-mode: upright;
    -ms-writing-mode: upright;
    /* Mozilla/5.0 (Macintosh; Intel Mac OS X 10_6_8) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/30.0.1599.69 Safari/537.36 */
    text-justify: inter-ideograph;

    As far as I know it works.

  5. Victor Mair said,

    May 3, 2017 @ 12:46 am

    Here's an interesting sidenote about using dictionary serial numbers to designate Chinese characters.

    In the latter part of the sixties and throughout the seventies when I was most intensively engaged in Buddhist Studies, several of my mentors and later colleagues who were excellent Buddhologists but who didn't know Chinese had a unique way to write Chinese Buddhist technical terms with numbers. Since they didn't know how to recognize, write, or pronounce Chinese characters, they would refer to them by their Mathews' Chinese-English Dictionary numbers. Instead of pronouncing the characters or writing them, these scholars would read off or write a succession of Mathews' numbers. They were experts in Sanskrit, Pali, Prakrit, Tibetan, Sogdian, Khotanese, etc.; many of them were extremely distinguished in their field. Some of them actually — through long use — memorized hundreds of the Mathews' numbers. It was uncanny. I remember one of the most distinguished among them, Edward Conze, sometimes in his lectures to large classes reading off Mathews' numbers, and we students would dutifully copy them down to look up later. It reminded me of Chinese telegraph operators who also memorized hundreds of arbitrary numbers to send and receive telegrams.

  6. Bathrobe said,

    May 3, 2017 @ 1:09 am


    I wasn't too taken aback.

    The first problem is the very modern assumption that if it's "Chinese" the Chinese must do it better. This isn't necessarily true, especially given that Chinese characters are also part of the Japanese and Korean patrimony.

    If you look at Wikipedia you'll find that "Tetsuji Morohashi was originally motivated to create a dictionary in 1917 when he went to China to study Chinese. Trying to look up words in the largest available Chinese dictionaries was frustrating". The original Dai Kan-Wa Jiten appeared in 1955–1960.

    The Chinese answer to this, Hanyu Da Cidian, was published between 1986 and 1994. I have this dictionary, and I can assure you that, despite being compiled by "a team of over 300 scholars and lexicographers", it has important weaknesses.

    The second thing to be noticed is the dates. The HJAS acquired its fonts in the early 70s, when China was in the throes of the Cultural Revolution. While I'm sure that it might have been possible to source the fonts in Taiwan, it is likely that the Japanese had a better organised, more attractive offer. They are rather good at that sort of thing.

    By the time that the HJAS went electronic it was presumably the 90s. While I'm sure China had made enormous strides by then, was still a pretty rough-and-ready sort of place. I'm not sure that they would have been able to offer the HJAS what it wanted.

    No doubt there is someone who could provide more detail than this, but I think that is probably it in a nutshell.

  7. Victor Mair said,

    May 3, 2017 @ 1:22 am


    Why did they use Morohashi numbers for the missing characters, why did the font come from Japan, and why was HJAS later typeset in Korea? Three simple reasons:

    1. Morohashi was the best Chinese dictionary in the world, especially for classical studies.

    2. Both countries, Japan and Korea, had very long and very strong traditions of Sinology.

    2. China was cut off from the world by a Bamboo Curtain from the 1950s through the 1960s. It was only in the 1980s that scholarly communication between the West and China gradually became regularized. At the same time, because of the enormous emphasis on Marxism-Maoism and the attacks on the feudal past that were in vogue, traditional learning seriously waned during that period. Consequently, many Western scholars training to become Sinologists in the generation just before mine went to Japan to study, and their spoken Japanese was usually much better than their spoken Mandarin (if they had any at all). That emphasis on Japanese Sinology carried through them to many of their students who were in my generation or even slightly later. It was only in the latter part of the 80s and the 90s when increasing numbers of aspiring Sinologists went to China that Mandarin skills noticeably improved. Of course, after Chiang Kai-shek's government became stabilized on Taiwan during the latter part of the 50s and the 60s, there were opportunities to learn Mandarin from Mainland refugees there (I learned much of my Mandarin in Taiwan during 1970-72), but the numbers were relatively small.

    I hope these explanations make you feel less taken aback now.

  8. AntC said,

    May 3, 2017 @ 2:03 am

    Thanks @Victor, @Bathrobe. Yes that explains it.

    But I also asked about typesetters amongst the expat Chinese community in America. Weren't there Chinese language newspapers? And therefore typesetting machines?

    Or would they have too small a repertoire of characters for academic Sinology?

  9. Carl said,

    May 3, 2017 @ 7:40 am

    Looks like support for "writing-mode" in CSS has increased while I wasn't paying attention, up to 94% of users: Thanks for the tip. It is still hidden behind vendor prefixes, but that's easy to work around.

  10. Victor Mair said,

    May 3, 2017 @ 8:23 am


    "Or would they have too small a repertoire of characters for academic Sinology?"

    You rhetorically and intelligently answered your own question.

  11. Victor Mair said,

    May 3, 2017 @ 9:51 am

    Well written, Bathrobe!

    Two different tracks, indeed!

    Many outstanding Sinologists in the generations before me could read Literary Sinitic / Classical Chinese extremely well, but could barely speak a word of Mandarin or other topolect. Mine was the first generation of Sinologists for whom it was more or less expected that you were supposed to learn to speak Mandarin (or other topolect) reasonably well.

    On the other hand, for centuries missionaries and other Old China Hands spoke — sometimes flawlessly — one or another of the Sinitic topolects (very often NOT Mandarin), though some of them became highly accomplished in written Chinese at the same time (e.g., the early Jesuits like Matteo Ricci [1552-1610] and Nicolas Trigault [1577-1628]).

    During my long flights to and from Hong Kong last week, I finally found the time to go forward with the tabulation of the results on the survey of easy and hard languages that we carried out on LLog starting from March 4. (I get more solid work done while sitting on a plane at 35,000 feet in the sky than when my feet are on terra firma –far fewer interruptions and diversions!) I'm happy to report that I will finally be able to present my findings in a post within one week. Not too surprisingly, data from the survey — in which more than 150 people participated — do indicate a sharp division between spoken and written with regard to perceptions of difficulty that they present.

  12. Victor Mair said,

    May 3, 2017 @ 9:55 am

    Further on the key role of Japanese in Sinology, the great Burton Watson, to whom we recently paid tribute on Language Log, was far more at home in Japanese than in Mandarin.

    I should also note that all of the best programs in Sinology at major universities in the United States, and probably elsewhere as well, require extensive training in Japanese language because of the high value of Japanese Sinological scholarship.

  13. liuyao said,

    May 3, 2017 @ 11:56 am

    Very interesting anecdotes.

    Both the Japanese and the Korean call the "Chinese" characters in their writing hanzi 漢字 (but read kanji and hanja), the reference being to the Han dynasty. It's somewhat like having English or German typsetters for the Roman alphabets.

  14. AntC said,

    May 3, 2017 @ 7:25 pm

    Thank you @Victor for the further reply.

    @liuyao It's somewhat like having English or German typsetters for the Roman alphabets.

    Yes Caxton employed German typesetters at first for printing English. Which has added to the mess of English orthography. Hence my question about non-Chinese typesetters.

RSS feed for comments on this post