The point is that it's not possible to neatly separate the two.I'd argue that any subset of English that can be parsed deterministically is a formal language, not a natural one. COBOL and AppleScript _look_ more like English, but they don't have any more ambiguity, really, than C. (I use "deterministically" mostly to exclude ML approaches and other NLP techniques, which can get a lot out of natural English, but have probabilistic results.)
You tell me,
"If x is equal to two then set y to seven."
is that formal language or natural? Because it is parseable one way or the other, and it's also exactly the sort of way you'd describe an algorithm expressively (which means it is protected speech).
You could argue that you don't even have to go that far. Back when Bernstein v. US was being litigated someone printed a literal book with the PGP source code in it as a means to export it from the US. A copy of the book was purchased, flown to Europe, then the binding was cut and the pages run through OCR. I don't think it was ever litigated, but at the time it was argued that even source code printed in a book was protected speech. The Clinton administration mooted the case by granting open source software an exemption from ITAR after the 9th circuit ruled (but before it could be heard en banc) that at least source code was speech.
Last edited: