There is a research and industry effort underway to replace TCP/IP with content-centric or named-data networking. No more servers or IP addresses. All content would be uniquely identified and optionally signed. It would become much easier to control content distribution, since routers would be aware of content names and caching. Think Cloudflare writ large.
Thanks to both of you for the links. Although it's an interesting premise, I think we can safely assume any large-scale practical implementations of this kind of technology are some way off, so unfortunately I don't think it changes my initial conclusion: we're likely to see more locking down and partitioning of today's Internet for a while, before any more open and robust long-term solutions take hold (perhaps based on the kind of alternative future architectures you mentioned).
