which moves (encrypted) fragments of files around the world, ostensibly for performance and reliability reasons.
So it would act like a content delivery network does with whole files.
Except that this layer would be the default assumption for where you put data on the Internet.
Data in the new paradigm has no home physical location. It only has identity, and access rights granted by possession of decryption keys.
For data intended to be fully public, perhaps its metadata would be unencrypted in the layer, for searchability. But that would not imply a particular physical location for the data file payload itself. A search would result only in an identifier, which the layer infrastructure would locate an retrieve from multiple sources.
Data would automatically maintain sufficient worldwide distributed copies of itself, and the system would migrate (and cache) copies of data fragments closer to end-users of the data, based on speculative probabilistic co-access patterns. In other words, data would coalesce toward where it was needed, as an automagic feature of the distributed storage layer.
This kind of distributed encrypted storage layer thing (not owned by any single company of course, but rather both open/libre and partly peer-to-peer) needs to get implemented, and widely adopted so that it is a default assumption of how content on the Internet mostly works, BEFORE it is made substantially illegal by overreaching governments.
That's how to make the Internet remain borderless. Make it a fait accompli that is very hard to subvert technically without blocking nearly every ip address, which, if this is implemented right, could be a partial mirror of fragments of the content.