Oddbean new post about login | logout I gave up trying to understand how their stuff actually works.
Looks like a copy of https://zeronet.io , that never really took off but was a neat idea
I guess there was no killer application for it: People who really needed censorship resistant websites also needed privacy, which bittorrent doesn't offer.
You could put an onion address in every torrent file
it is easy to think that we are using Bittorrent, since we are using its Mainline DHT, but we are Not. No p2p storage at all. Just a good old web server, but the censorship resistance come from the fact that you can always point your public key to another hosting provider if you got censored or deplatformed. In fact, because it is just DNS packets over Mainline, you can use SVCB records to point to mirrors of your data, so even if your main host is taken down, http agents can be smart enough and failover to reading from your configured mirrors.
Very interesting. If not using BitTorrent's DHT what is Melvin's talk about millions of nodes referring to? Webservers? DNS servers?
These are routing nodes, they have in-memory storage and they only store two things: 1. routing table to nodes closer to a target info hash (basically a hash table automatically sharded) 2. small packets of arbitrary (and other non arbitrary but less relevant to Pkarr) data in an LRU cache. Nodes churn all the time and even when they don't churn if their cache is full your data might be evicted. so it is great for censorship resistant routing, not so much for data durability let alone storage of large blobs.
Ok so the DHT is the BitTorrent mainnet but it's only used for small signed DNS packets. What are the chances BT clients filter out these as unwanted traffic?
BEP0044 is a standard, nodes are free not to support it. It is for arbitrary data, our traffic is not any different from any other traffic, if we abuse the network we will get rate limited like any other spammed. That being said we are doing our best to avoid abusing the network, by heavily caching packets with large TTL values. There is a possibility that nodes might filter out packets that look like DNS, but first they don't have incentive for that, we are not hammering them or costing them much, second that will require a software updated distributed to millions of clients, not very likely, in fact networks like this have the opposite problem: legacy.
We ARE using the DHT, but we are not using the Torrent aspect. These are two orthogonal parts of Bittorrent. You can use Torrent swarms without DHT (using trackers) and you can use the DHT without using it for connecting to peers and sharing content, you can use it to just advertise small arbitrary packets. which is why Pkarr is possible at all. you can read more at BEP0044