Occasionally, in the course of using the Internet, one finds reason to ask, “How much can ISPs rewrite content without users noticing?”
Firstly, I have never seen a page whose design had any errors to speak of. Is there a technological implementation of this or some unknown somehow non-technological celestial one? I theorize that automatically correcting all design errors requires at most the intelligence level of an intelligent human, and if an AI can be implemented at that level, then it can be pointed at any website or page and fix any errors. But human level AIs have not been observed yet.
Instead, the following is true: any website can be automatically rewritten with less than human-level intelligence based on a parse of its content using, e.g., a combinatory categorical grammar library with semantics and a semantic analysis of the HTML structure of the website or a test rendering of the page using a browser engine with a “vision agent” or equivalent rendering test logic. In theory, both the design and content of any website would be able to be rewritten by the ISP to eliminate all errors if the ISP has access to all files/pages on the site at once, which is theoretically possible with either shared hosting or VPSes. In theory, an ISP could run circles around a user by modifying their content without telling them, with the only problem being conversations with other people the user has that include links to the user’s content. In addition, any JavaScript published by the user would theoretically be able to be buffered, cached, and dynamically modified to work with a rewritten page structure or content through deterministic semantic analysis of the original script source.
For conversations that include links, it may be theoretically possible to alter messages on-demand based on the original and modified content, which would only become a problem if the users discuss the problem in meat-space; however, there is no possible solution to that.