Viewing a document on the web needs to be decidable. The original design of the web was HTML documents with forms. This IBM 3270 style design used the browser as the user interface to server-side programs. The browser's job of presenting a document was decidable and the form submission, page load process allowed the user to understand and control the data sent to a server. The server learns what was in the form and URL when the user decided to click the submit button.

Moving the software into the browser improved latency, but questions like "Is this webpage doing something dangerous/annoying?" became undecidable. If we provably cannot determine if a webpage will halt without running it, we obviously cannot as any more complicated questions about the webpage's behavior. As long as webpages have access to a Turing complete language, the browser will be de facto an OS. Unfortunately, returning to a web of documents isn't going to happen anytime soon; too many people profit from this ability to run programs on other people's computers.

(Gary Bernhardt's amazing (and terrifying) talk "The Birth & Death of JavaScript"[1] was absurdist comedy, not a guide to future browser designs)

[1] https://www.destroyallsoftware.com/talks/the-birth-and-death...