Every bitcoin node has a full copy of the blockchain, which is why it's so important to keep block sizes low. Nostr relays won't possibly be able to store a copy of the full multi-zetabyte network at scale (and don't have a consensus protocol anyway)
Discussion
Yes that's why I said in a way
Follow a protocol so to speak. I didn't mean every relay store everything. Perhaps nip1, the only required nip, should have provisions for outbox model for clients if this is how we think the protocol should work.
in theory, relays could become clients to other relays and forward requests
i have an interface for this kind of multi-level event store interface built into my #realy event store library, i have implemented a shitcoin layer 2, and two things i have penciled in for the future is a blossom store that can only search for event IDs (first layer must contain indexes) and a second model is where relays use each other as event stores and forward queries to other relays and cache the results for the relay's users
these are two distribution mechanisms i have in mind for how nostr scales a lot further, i believe there is similar kinds of multi-node query forwarding design in blossom too
ultimately this is a massive privacy and bandwidth benefit because one can rent a few good, reliable, fast, local relays to do your relay work and they automatically sync up with what you want to see from other relays but they make the requests on their behalf, which means you don't get seen doing it, and you don't waste your bandwidth doing it, and because it caches, it doesn't have to be done again if other users on the same relay want to see that content