Nuh 🔻 on Nostr: 1. Because it only works when we have common relays, these relays have to be able to ...
1. Because it only works when we have common relays, these relays have to be able to serve the entire network, if they can't then most of the network becomes degenerate isolated bubbles.. so yes it works well, but can it scale to be the next web?
2.Mainline bootstrap nodes are NOT a form of centralisation, I can build an app and bundle it with my own node as a bootstraping node, and another developer can hardcode their own bootstrap nodes, and a third app might keep storing their routing table to disk and use it for bootstrap in next sessions and ALL these apps will function normally. You don't need THE bootstrap nodes, you node ANY bootstrap nodes. In fact it is quite difficult to make a DHT network separate from Mainline if it is compatible.. because only one dude can force both networks to merge (which happened before) so it is not a centralisation point at all.
3. Mainline has around 10 million routing nodes, at least 1 million of them support storing small arbitrary data, it is not encrypted, it is extremely reliable, and I don't think millions of small nodes have less capacity than Damus. Even if so, as long as it offers the ultimate censorship resistant fallback, we can, should and do use caching layers just like DNS scales.
4. Using Pkarr is as easy as `to_socket_addrs` more importantly it requires just as little context, debugging that comes down to network connectivity, where as there is a whole social graph and happenstance that is involved in Outbox model that will simply limit the reliability of the system at scale. I claim there is much less learning curve when your system api and behaviour is trying to stay close to what you expect from DNS
5. Pushing in the DHT direction is only meaningful if you are willing to use Mainline which comes with dramatic change to ed25519 keys as the root of the identity. Some keep saying, if we felt like it we can make a secp DHT, and that is just ignoring how difficult it is to bootstrap an overlay network from scratch and ignoring how much size really matters if your purpose is censorship resistance... if you are going to end up with a few hundred nodes (which I bet would be the order of magnitude from watching other popular DHTs that don't have Mainline product market fit) then I claim you are not going to be in any better position than you are currently at.
2.Mainline bootstrap nodes are NOT a form of centralisation, I can build an app and bundle it with my own node as a bootstraping node, and another developer can hardcode their own bootstrap nodes, and a third app might keep storing their routing table to disk and use it for bootstrap in next sessions and ALL these apps will function normally. You don't need THE bootstrap nodes, you node ANY bootstrap nodes. In fact it is quite difficult to make a DHT network separate from Mainline if it is compatible.. because only one dude can force both networks to merge (which happened before) so it is not a centralisation point at all.
3. Mainline has around 10 million routing nodes, at least 1 million of them support storing small arbitrary data, it is not encrypted, it is extremely reliable, and I don't think millions of small nodes have less capacity than Damus. Even if so, as long as it offers the ultimate censorship resistant fallback, we can, should and do use caching layers just like DNS scales.
4. Using Pkarr is as easy as `to_socket_addrs` more importantly it requires just as little context, debugging that comes down to network connectivity, where as there is a whole social graph and happenstance that is involved in Outbox model that will simply limit the reliability of the system at scale. I claim there is much less learning curve when your system api and behaviour is trying to stay close to what you expect from DNS
5. Pushing in the DHT direction is only meaningful if you are willing to use Mainline which comes with dramatic change to ed25519 keys as the root of the identity. Some keep saying, if we felt like it we can make a secp DHT, and that is just ignoring how difficult it is to bootstrap an overlay network from scratch and ignoring how much size really matters if your purpose is censorship resistance... if you are going to end up with a few hundred nodes (which I bet would be the order of magnitude from watching other popular DHTs that don't have Mainline product market fit) then I claim you are not going to be in any better position than you are currently at.