I am afraid that just like computers that one mind could completely understand, networks that one mind can completely understand are a thing of the past. This is quite sad, because, as the author noted, they were a great way to learn
> The result is a system with properties that work with today's internet to give us the pleasant, simple programming environment of the '90s LAN:
> -Use the global internet identity system of your choice for authentication, and do cryptographic authorization at the IP level.
>- Keys are generated and rotated for you automatically.
>- People map directly to unspoofable IP addresses.
>- Run custom servers on your network and access is limited to only those people on the network.
>- Your data is protected by the simple yet powerful social dynamics of small groups.
This is exactly what libp2p (https://docs.libp2p.io/introduction/what-is-libp2p/) is trying to be: The network isn't defined by IPs but by cryptographic identities, routing is done directly through those identities, end-to-end security is built-in and communication is independent from transport, which can be absolutely anything.
My guess is a few decades, if it can hold until then. libp2p is a huge stack that redefines a lot of what we're doing today, and reimplementing the whole thing is a monstrous task. But if it could become mainstream that would be a very good thing.
