How exactly are you going to "bypass BGP" on the global Internet? Reaching your trusted peers depends on routing, which means BGP (at least for anything outside of your ASN.)
Well if you want to get rid of BGP then it probably doesn't make sense to talk about ASNs since BGP is a way to connect ASNs... That being said, it does seem very pie-in-the-sky to imagine the entire global internet using decentralized routing without some sort of backbone infra that is at least moderately centralized. There is just too much data. I could imagine mesh networks being more feasible if we were still living in an era of html websites and RSS feeds.
All data should only exist within the light cone that it is used within. That means a lot of data is improperly used today in terms of where and how it is moved.
You need a lot less information "over the wire" when it doesn't move as far, and when you can speculate and predict what to render (ML/AI). What matters is the package and space more than anything. Go figure. That's also why "massively distributed compute/swarm computation" makes sense in this mesh like model you hint at.
The future is strictly local data unless we totally escape time dialation over large distances, whether or not we bypass speed of light limits.
The future is strictly a simulation if we are a multi-galaxy species, let alone spread across the galaxy. The only way to have a real-time conversation is to simulate the person/ego. The local copies would be as real/eventually consistent as possible over time.
Maybe it is too much data for devices that have to share a channel. But lasers (think: old TV remotes using infrared) do not have to share channels with each other, and can be very high-bandwidth and more difficult to snoop on indeed. Mirrors could even be used for relaying, to save lots of compute. Like fiber cables piercing through the air itself!
I was speculating that the notion of a laser-based mesh network could deserve more attention, if it is economically viable. That seems like a better choice than something like WiFi, because you do not have to share a channel with huge amounts of other devices.
Several kilometers of range seems tolerable with enough collaboration. I do wonder though, whether mesh networks for bidirectional (synchronous) comms are a good idea in hostile environments. It seems too easy to map out the positions of the nodes. Systems that support distrbuted, asynchronous comms and use of wireless protocols on phones seem like a good match for them though.
