Nuh 🔻 on Nostr: Thanks for referencing Whanau, I don't have a chance yet to read it in depth, but I ...
Thanks for referencing Whanau, I don't have a chance yet to read it in depth, but I got enough idea to make a judgment.
1. It is way more complex than BEP0005, which wouldn't be a deal breaker if it actually delivers.
2. It assumes one of the hardest thing to assume ever: a web of trust! Its innovation seems to be leveraging such social graph efficiently but that is ignoring the fact that a robust and decentralised WoT is the holly grail of the Internet and sybil resistance in general, you can't just take it for granted.
3. In practice, trying to bootstrap this social network is going to directly translate into even smaller and slower growing DHT, because instead of just running a Bittorrent client that uses upnp to rund an embedded DHT, you need users to manually set up what nodes they trust ... that would at least lower the realistic size of your DHT in the foreseeable future by many orders of magnitude... for comparison Hyperswarm DHT that ONLY requires a public IP that isn't behind NAT, still never managed to have more than 2000 nodes.
4. I worry that the reality is going to be worse, and you will have a hub and spoke model where the social graph doesn't just uniformly churn, but can take sudden big blows whenever a Damus-like node churns, again this is more likely because unlike normal DHTs you can't just start trusting a hardcoded node, then start trusting others, that defeats the purpose, you need explicit user input in establishing trust edges
5. So yes you might be sybil resistant, but if that comes at the cost of being not only extremely small but also much harder to grow (social scalability), then you already lost, because you still need to serve the entire Web AND the DDoS.
If I am correct and a social network based DHT is impractical to scale beyond few hundreds of nodes, then that is not any different from few hundreds of nip65 only Nostr Relays that all keep the entire list of each other in memory.
What I am trying to say is, if you put even more friction in the way of scale, then whether your network is structured (DHT) or unstructured (flat list of gossip peers) is irrelevant, both are just as vulnerable to DDoS and can't scale as much as a DHT of millions of nodes.
That being said, ISPs already have this trust graph, so it would be technically interesting if all the world ISPs turned into a DHT :)
If you managed to pull this impossible task though and you built DHT that is both sybil resistant AND growing fast and consistently enough that it has a chance to resist DDoS, you will be my hero, especially if it supports ed25519 as well as secp, so that we can use it besides Mainline.
1. It is way more complex than BEP0005, which wouldn't be a deal breaker if it actually delivers.
2. It assumes one of the hardest thing to assume ever: a web of trust! Its innovation seems to be leveraging such social graph efficiently but that is ignoring the fact that a robust and decentralised WoT is the holly grail of the Internet and sybil resistance in general, you can't just take it for granted.
3. In practice, trying to bootstrap this social network is going to directly translate into even smaller and slower growing DHT, because instead of just running a Bittorrent client that uses upnp to rund an embedded DHT, you need users to manually set up what nodes they trust ... that would at least lower the realistic size of your DHT in the foreseeable future by many orders of magnitude... for comparison Hyperswarm DHT that ONLY requires a public IP that isn't behind NAT, still never managed to have more than 2000 nodes.
4. I worry that the reality is going to be worse, and you will have a hub and spoke model where the social graph doesn't just uniformly churn, but can take sudden big blows whenever a Damus-like node churns, again this is more likely because unlike normal DHTs you can't just start trusting a hardcoded node, then start trusting others, that defeats the purpose, you need explicit user input in establishing trust edges
5. So yes you might be sybil resistant, but if that comes at the cost of being not only extremely small but also much harder to grow (social scalability), then you already lost, because you still need to serve the entire Web AND the DDoS.
If I am correct and a social network based DHT is impractical to scale beyond few hundreds of nodes, then that is not any different from few hundreds of nip65 only Nostr Relays that all keep the entire list of each other in memory.
What I am trying to say is, if you put even more friction in the way of scale, then whether your network is structured (DHT) or unstructured (flat list of gossip peers) is irrelevant, both are just as vulnerable to DDoS and can't scale as much as a DHT of millions of nodes.
That being said, ISPs already have this trust graph, so it would be technically interesting if all the world ISPs turned into a DHT :)
If you managed to pull this impossible task though and you built DHT that is both sybil resistant AND growing fast and consistently enough that it has a chance to resist DDoS, you will be my hero, especially if it supports ed25519 as well as secp, so that we can use it besides Mainline.