[olliej]

Systems like you're describing, exist, and the fundamental problem is that they are not even semi-permanent archives.

For something to exist, someone has to host it, and the way you get something hosted is to pay for it (either paying someone to be a host, or paying for hardware and connectivity). Once you stop paying those costs you're reliant on other people choosing to keep your data around, just as archive.org does. If no one chooses to, the fact that you had a pile of random hashes scattered into the resource naming/identification scheme does not matter. Sure nodes would cache commonly accessed data, but the moment it stops being frequently used it starts getting pushed out of those caches to hold the new popular stuff. If you are hosting it yourself, or paying someone else to host it, once it drops off the "being popular" wagon its persistence is limited to whenever the next cache flush occurs.

So in exchange for content being harder to update, the routing performance being lower, making cryptography impossible, not working for dynamic content, and making censorship much easier, you have not solved the problem that archive.org already attempts to solve. Nothing in your scheme would obviate the need for scraping and separately archiving, nothing ensures content remains once no one is paying to ensure hosting.