I still think we need a relay indexing service: a relay whose sole purpose is to know where each event can be found.
Discussion
That’d be very expensive to run ergo centralizing. Sure, that’d be faster/more effective than hints + gossip, but at what cost? I think gossip and more hints can get us there
The relay doesn't store any event. It just store pointers to where they are. The idea is to decentralize. Today there are 3000 relays available. We only use 300 or so.
Yeah, but that relay still would need to keep all event ids; I haven’t ran the numbers but this sounds expensive
A service discovery would be a better name i guess, but then you are nor truely decentralized, you will depend on one single node (or bunch of replicas) to function. If that service discovery node is blocked, you can't do anything. Except:
1-service discovery is optional and only to accelerate fetches.
2. every node is a service discovery node, which they keep themselves updated with a bunch of master service discovery node. Making the clients less dependent on master service discovery nodes and even being able to not expose them.
It is a service discovery. Anyone could run it. It's not centralizing.
I suspect we would do well to go back and take a look at WAIS, which was the road not taken for Internet content when centralized search engines and directories won out. Just as nostr is a lot like a better Usenet, maybe we can steal some ideas for creating decentralized search?
As long as anyone can run that service, it should be better.
Now we have zaps, why is there no mechanism to set a budget and autopay for relays so they can have a revenue stream?
We just need a NIP for that. nostr:npub1qqqqqqyz0la2jjl752yv8h7wgs3v098mh9nztd4nr6gynaef6uqqt0n47m, any thoughts?
Wow good question. I’ll think about this
DHT? In this case it's more like CHT😂