Ever since I read about RINA [1] some ten-odd years ago, I've always thought that this was the future standard of the Internet. I kept track of some implementations [2][3][4] but they've since been overtaken by developments in TCP/IP, most notable (for me) among which is QUIC.

I've come to believe that rather than starting from a clean slate, we need to work within TCP/IP's constraints to bring us piecemeal from "rough consensus and running code" to sound networking theory.

Zocket is my contribution to this effect. It's what you get when you marry a sockets API to zero-configuration networking--specify apps by their own names rather than their host's addresses. It's pretty rough, but the happy path mostly works.

I can imagine use cases for this in cluster networking, especially in the cloud. But I'll probably start with something smaller, like how to backup pictures from a smartphone to a home server despite the latter being configured with DHCP.

Happy April Fools' Day!

[1] https://www.martingeddes.com/think-tank/network-architecture... [2] https://github.com/IRATI/stack [3] https://github.com/rlite/rlite [4] https://github.com/dstaesse/ouroboros