Fable
The increasingly economized type molds in the print shop
The academy wanted to copy ancient texts from various regions into uniform booklets. The copyists got stuck on rare characters and wrote too slowly on long words. The director wanted to make a symbol table so that short strokes could also spell out long sentences.
At first, the director had everyone use one character per cell. Common characters were easy, but out-of-town place names and newly coined words could not fit. Someone proposed putting every whole word into the table, but the resulting volume was too thick to move, and copyists turned pages more slowly than they copied.
The young copyist came up with a compromise. He first broke all sentences into the smallest strokes and character blocks, then counted which adjacent small blocks most often appeared together. If 'clear' and 'spring' were always next to each other, he merged them into a new block; if 'east' and 'gate' often appeared, he merged those too.
The first revision worked well, and writing "Qingquan Street East Gate" took many fewer strokes. The director then urged him to merge all common combinations into it. Before long, the symbol table began to grow fat again, with some combinations that had appeared only a few times taking up space, while the truly useful blocks became harder to find.
The copyist changed to merging round by round. Each round picked only the pair that most often appeared side by side, wrote it into the register, and recounted. In this way, common words naturally grew into larger blocks, while rare words could still be assembled from small blocks.
A student asked why they didn't just bring the whole dictionary from the start. The copyist spread out two sheets: one was too fragmented, and everyone wrote slowly; one was too full, and everyone searched slowly. A good ledger must sit between fragmented and whole, leaving room to piece things together.
The director then tested with foreign names. If the symbol table only included common local words, foreign names would shatter into fine fragments; if every foreign name were included, the table would swell. Progressive merging gave the academy a way to handle both the familiar and the unfamiliar. The copyists also set a capacity for the table. Merge too little, and everyone's hands still ached from writing; merge too much, and looking up the table became slow, and rare combinations could be mistaken for common ones. Each time a new block was added, they had to see how much copying it saved and how many table slots it took—whether it was worth keeping. The copyists also found that the table wasn't fixed for life. The more new books there were, the more common combinations changed; old combinations rarely used again also had to give up their place.
Later, merchants from out of town brought new place names, and the academy did not need to rebuild the entire symbol set. The copyists broke unfamiliar names into existing small pieces, combining them where possible and gradually merging frequently occurring ones. The booklet was copied faster and faster, yet did not lose its flexibility to accommodate new words.