DHT is a perfect fit, as the relays are the endpoints.
"Gossip" is just a happytalk way of dodging the question.
How do you know which IP addresses to gossip with? "All of them" doesn't scale. A DHT is nothing more than gossip with a distance metric so you can know when you're getting closer or farther to your goal.
I note that you posted no links. A lot of this stuff sounds great in quick replies but doesn't pan out when you have to sit down, write, and implement a detailed spec.
I don't think you understand how nostr works. The relays are not meant to be endpoints, they are only meant to be dumb relays. Also the relays already operate as a hash table for the notes being stored.
There's no IP addresses with nostr gossip, just a list of relays and some overlap between clients. If our clients share a single preferred relay, I can retrieve your note and rebroadcast it to my other preferred relays. It's not a hard concept to understand, and the "links" are on the nostr github.
reply
The relays are not meant to be endpoints, they are only meant to be dumb relays.
Yes, they are not meant to be endpoints, they are only meant to be dumb relays, but is this currently the case?
If most connect to the same big relays, is it really irrelevant which relay you choose? Are relays really fungible? Can you just spin up a new relay and have the same network effects as other relays?
Serious questions, I don't know the answer.
reply
by some axiom a relay shouldn't generally be treated as an end-point, hence the word "relay"
reply