Braydon Fuller on Nostr: I think the largest problem with a DHT is that it is primarily distributed by hash ...
I think the largest problem with a DHT is that it is primarily distributed by hash rather than a system of which can be based upon value—in short, it's inhuman.
Value is subjective, determined at the margin and always changing. For this reason trackers have been necessary in bittorrent, magnet links and etc. and from my understanding the inspiration for relays in Nostr. If trackers are necessary, a DHT seems like a complicated means of peer-to-peer sharing of metadata, without gain.
In Nostr, I think this means that relays should sync more with each other for a variety of purposes based upon value determinations. This can be syncing relevant notes based upon a social graph and to crawl and fetch all referenced notes therein; something that I am mostly interested. It can be to sync profile metadata and relay list metadata in a social graph that is multiple levels into a social graph for greater redundancy. It can be syncing and filtering notes based upon LLM determined topic relevancy. It can be more sophisticated algorithm based on various value scoring mechanisms. There are many possibilities and I think it was a good choice for the protocol to not have a peer-to-peer topology at the base level, various network topology can be added subjectively later.
Value is subjective, determined at the margin and always changing. For this reason trackers have been necessary in bittorrent, magnet links and etc. and from my understanding the inspiration for relays in Nostr. If trackers are necessary, a DHT seems like a complicated means of peer-to-peer sharing of metadata, without gain.
In Nostr, I think this means that relays should sync more with each other for a variety of purposes based upon value determinations. This can be syncing relevant notes based upon a social graph and to crawl and fetch all referenced notes therein; something that I am mostly interested. It can be to sync profile metadata and relay list metadata in a social graph that is multiple levels into a social graph for greater redundancy. It can be syncing and filtering notes based upon LLM determined topic relevancy. It can be more sophisticated algorithm based on various value scoring mechanisms. There are many possibilities and I think it was a good choice for the protocol to not have a peer-to-peer topology at the base level, various network topology can be added subjectively later.