Writing With Unnatural Constraints

Marcin Wichary, writing at Unsung, regarding a rather subtly astonishing FAQ for the game Super Metroid, written in the late 1990s by “rs1n”. It’s over 17,000 words long, and the whole thing is full-justified when viewed with a monospaced font, using no trickery with extra spaces or hyphenation to achieve the layout:

The author lightly covers in the FAQ at the bottom:

What program did you use to justify the text?

None. I just chose words carefully so that everything lined up on the right hand side. Everything was done with an ASCII editor.

I’m sharing this mostly as a curiosity; some rewrites for physical books are par for the course to avoid widows and orphans, but you don’t see them as much in onscreen writing.

The FAQ reads pretty well, but to my ear, there’s something obviously off about it. The justification gimmick is very fun and the fact that the prose reads as well as it does is quite the feather in rs1n’s cap. He tried something very difficult and pulled it off with aplomb. And I’m sure many people have read the FAQ without even noticing the accomplishment, or considering just how difficult it is to write with such an unyielding constraint. There were many FAQs and other documents from the plain-text era that were full-justified, using extra spaces between words and paragraphs and occasional hyphenation to achieve the effect. rs1n’s FAQ achieves the effect with no trickery, and the result looks so natural that it doesn’t call attention to itself. It just looks nice.

But it doesn’t read entirely naturally. It’s stilted, as though written by someone for whom English is their second language. Which, for all I know, is true for rs1n, but I doubt it — I think the ESL-like stiltedness is just the result of making word choices to fit the line length, not simply for meaning and tone.

The line-length constraint brings to mind lipograms, “writing paragraphs or longer works in which a particular letter or group of letters is avoided”. Ernest Vincent Wright published an entire novel in 1939, Gadsby, without using the letter e. The Wikipedia page for lipograms lists more examples. That’s a hell of a trick, but there’s a reason why Gadsby is almost unheard-of, and Fitzgerald’s The Great Gatsby remains famous. Gadsby is miserable to read. Wright was obviously a fine writer. Gadsby’s “Introduction” is written without the no-e’s constraint — and largely about the no-e’s constraint — and it reads just fine. But the text of the novel is a goddamn laborious slog. I couldn’t get through more than two paragraphs before deciding that my point was proven. It’s a fun challenge to compose lipograms but it’s no way to actually write. The gimmick necessarily becomes the highest priority of your word choices, not clarity or meaning.

So too, goes my thinking, regarding the “watermarking” of AI-generated text through word choices alone. Fitting words into the watermarking pattern surely results in prose far less stilted than omitting the letter e, and surely also less stilted than rs1n’s hard-wrapped-at-75-characters-per-line Super Metroid FAQ, but it’s necessarily stilted to some degree. It’s sophistry to argue that one can write within a constraint without the constraint affecting the quality of the writing.

Bonus Quibble

Wichary opens his post with the following:

I don’t think anyone particularly enjoys typesetting in monospace.

I disagree! I think typesetting in monospace is fun — or at least satisfying. I think it’s part of the appeal of Markdown that you can make a document look just right. What you see reading Markdown is what I saw writing it. And just look at the fun people still have printing “large type” banners using ASCII punctuation characters or fancier box-drawing Unicode glyphs. When you format something in plain text, you know exactly how it’s going to look everywhere.