From Kingdom of Characters: The Language Revolution That Made China Modern, by Jing Tsu (Riverhead Books, 2022), Kindle pp. 64-69:
Zhou’s breakthrough was figuring out how to correlate the physical act of selecting and retrieving a character in a tray with the mechanical motion of preparing a character, etched onto a cylinder, to be inked and printed. The result was a single coordination of mechanical motions, optimally economized and completed by a human operator. In his article for The Chinese Students’ Monthly, Zhou documented the details of his typewriter and the technical challenges he had to overcome. He had known, of course, that he was not the first. There was Sheffield’s machine and an earlier Japanese prototype, both of which he acknowledged as having arrived at similar ideas independent of his endeavor. It was a respectful nod to those who came before him.
Just as he was preparing to publish his article, though, an unexpected challenger nearly derailed Zhou’s debut. Qi Xuan was a fellow Chinese student studying engineering at New York University. Unbeknownst to Zhou, Qi had been on a parallel track. Relying on a different set of principles for building his own Chinese typewriting machine, he figured out how to do what Zhou, Sheffield, and others thought was impossible: arrange the characters by parts.
…
In place of Zhou’s character grid, the keyboard of Qi’s machine had only three keys—a back spacer, a forward spacer, and one that selected the character. But Qi’s machine also relied on cylinders of type (two cylinders instead of Zhou’s four). The upper cylinder, with characters inscribed on paper, served as a guide. The lower cylinder, with corresponding characters engraved on a copper surface, made the actual impression. An operator would use a hand wheel to rotate the machine’s upper cylinder until the correct row appeared in a viewfinder on the front of the machine. Then three keys would be used to select the correct character from the row and lock the cylinders in position, aligning the corresponding character cast on the lower cylinder and stamping it on the page.
Though they looked very different, the underlying design and mechanisms of Zhou’s and Qi’s machines were very similar. And Qi’s handled 4,200 actual Chinese characters, just 200 more than Zhou’s core lexicon. Qi’s machine, however, was different in one very important respect: he had broken with the prevailing commitment to reproduce whole, complete characters. Of his 4,200 individual Chinese characters, 1,720 were in the form of character components, radicals, and their possible variant positions in a square space, which allowed his machine to generate, in theory, more versions out of the same parts. In three steps, using these keys, the operator could purportedly produce 50,000 combinations.
Operating his machine, Qi explained, was closer to spelling an English word than producing a Chinese character. If you treat radicals like groups of letters, you can play with different combinations the way you would in a word game. Let’s say you have three English words—“exist,” “expect,” and “submit.” You can generate more words by mixing and matching their parts to form new words like “sum,” “suspect,” “subsist,” “bit,” “mist,” “its,” and “sex.” Unlike modern English words, which have equal spacing for each letter and line up in a neat row, components and radicals can move around from character to character. In print they can occupy different quadrants of the square space that each character fills, which means their position, and consequently their size, can change. For example, the character for “fire,” 火, fills an entire square space when standing alone, but it becomes thinner when it is a radical put on the left of the character “braise,” shao 燒. In turn, it changes form altogether—into four flames—when placed at the bottom of the character for “hot,” re 熱. Qi accounted for possible variant positions like these by giving them separate engravings on the cylinders. Consequently, there are at least three options for “fire” to be combined with other components and thereby form a greater number of characters.
Qi challenged the idea that characters had to be individual, stand-alone units. He thought of them as more modular, like alphabet-based words, things that could be recycled to compose different characters. And once he started tinkering, some of the Kangxi radicals did not sit well with him. He took more liberties and slipped in a few radicals he had devised that he thought worked better. A bigger shift than Zhou’s was creeping in. Others would also start to ask whether exclusively using radicals for character classification still made sense.
…
For different reasons, Sheffield and Zhou both concluded that characters were the way they were for good reasons and preserving them was of the utmost importance. Qi’s scheme “looked well on the face of it,” Zhou cautioned, “but they forgot, that the same ‘radical’ in different characters differ [sic] not only in size but also in shape, and, furthermore, they occupy different positions in a character.” He named one example, the square-shaped radical meaning “mouth” 口, which shows up in the characters for “ancient” 古 and “cry” 哭 but in each its size and location are quite different. Consequently, one cannot use the same square for “mouth” in both cases, as they were designed to fit into different configurations. If the different sizes of 口 were forced together, they would create absurd-looking characters with overlapping strokes in all the wrong places.
Zhou’s design eventually won out.
Japanese modernizers had a much easier time adapting katakana for use in typewriters, semaphore, and telegraphy.
Far Outliers 