I've never understood what Knuth considered literate programming. I've seen some examples of his and find them incomprehensivle. I think the issue with Knuth's contribution is that he only seemed to concentrate on the specification of a markup language, marred by TeX baggage.
The key idea of notebooks today are their interactiveness, the dynamic between a markup state and view state, and their multi-paradigm and multi-language nature.
Yeah, if you want to search for ancestors, I think literate programming definitely qualifies. I've also heard references to Smalltalk and Lisp machines as ancestors, but not being familiar with either, I can't say.
Knuth introduced the technique of literate programming in 1984, MathCAD was first released in 1986, and my copy of Literate Programming (the book of the thing) was published in 1992.
Knuth's literate programming notably differs from notebooks in that it was designed to write explanations meant for humans, which meant that the author could present the code in chunks in the narrative sequence; the code did not have to be in execution order, unlike notebooks.
I'm not clear on this. Is it not just about having the function definitions in an arbitrary order just like all of us mortals do? Is there anything special about the code structure that Knuth proposes?
No, Knuth's cweb has the notion of chunks, which is just lines of code. It wires the chunk in dependency order. A chunk can have multiple functions, or no functions, or parts of functions; cweb doesn't care.
Part of the motivation might have been that he set up WEB for Pascal, which does enforce an order of declarations. Once he started working more with C, perhaps there was less reason; but he found that he enjoyed working in that manner. I haven't looked at anything released in that format for years, though.
The key idea of notebooks today are their interactiveness, the dynamic between a markup state and view state, and their multi-paradigm and multi-language nature.