>>157>cjdnsIt's not bad, but:
- It's not an anonymous routing
- Has no concerns about leaking metadata
- Uses IPv6, that had some flaws in it's history (although it's encrypted by default on cjdns, which is good)
- Still relies on cable and not on EM waves such as DASH7
- Do nothing about the application layer, therefore, using the same soykaf we use normally: http
It's not bad, as I said, but I think a new network should be:
- Anonymous and distributed by default (maybe use steganography on the tranmission of the packets)
- Work as a meshnet, using something like DASH7
- Have it's own name resolution system, such as OpenAlias
- Leak very little metadata (the network should not give any UID to users... don't ask me how it would work, it's just an hypothesis)
- Use strong crypto, preference to PQ-crypto
- Have a formal proof and formal verification of the code (check Isabelle/HOL and Coq)
- Capability based, secure by default and simple code
- Have privsep by default