who runs the 3 dht on-ramps? is there a list?
Wow I just found nostr:npub13ndpm2hm9hud4azsq5euhf5mv3d05r90wymwxsd7rdn29609hhvqp60svh. He is so *smart* and not at all "retarded." π
I'm not sure if I should bother correcting his misunderstandings of nostr, or his misjudgements of what is and isn't censorable and why, or just ignore him. Does he listen or just talk? We will see.
I like pkarr. I like using Mainline DHT directly even more. I'm using it in Mosaic. But it is also censorable. The bittorrent DHT may not be subject to sybil attacks, but it only has 3 (AFAICT) public well-known on-ramps. Take those three down and people who don't already have a node list can't get on. Nostr is not better for bootstrapping (finding a person's relays) but it isn't much worse. There is a weak centralizing pressure to put relay lists on a the same well known popular relays, but those change over time and the pressure changes over time, and good clients republish to the current popular set which can move. The main issue is that client devs are all over the map in terms of how they choose to deal with the situation, and some aren't even using outbox model. IMHO (emphasis on H) the main benefit of the DHT is someone new with a single key they want to follow can get started without knowing anybody nor any relays/servers.
There is no censorability difference between homeservers in pkarr and relays in nostr. You can run your own, or you can outsource it, in both cases. One isn't more censorship resistant than the other.
When I look at https://github.com/pubky I see among top rpos "pubky-app" which says it is a deprecated repository, and will be replaced with pubky/franky. But pubky/franky does not exist. So IMHO there is no opensource pubky, which means I won't be using it and I recommend against.
Discussion
I don't know who runs them.
router.bittorrent.com:6881
dht.transmissionbt.com:6881
dht.libtorrent.org:25401
This is false, regardless of Nostr or Pubky or whatever low stakes feud bitcoiners engage in instead of trying to fix scaling issues.
I'm willing to admit that I am wrong and learn something. But just saying I'm wrong isn't going to do it. Like any kind of science, you have to show me that I am wrong.
Every node youβve ever talked gets cached.
Fresh installs can bootstrap:
- BEP-34 DNS seeds (TXT records listing live router IPs).
- Magnet links with explicit peers.
- Local peer exchange / LAN multicast.
- Manual seed lists shipped in the app (same trick Bitcoin Core uses).
Removing all these vectors simultaneously would require blanket UDP filtering across the Internet, which is a lot harder than blocking a handful of Nostr HTTP relays.
If these were necessary, why aren't they mentioned in BEP0005 or hardcoded in Libtorrent?
Also, take mainline crate, and replace the routing table with this node; relay.pkarr.org:6981 see what happens.
Also, once you bootstrap, call to_bootstrap() to cash your table to disk for next sessions, and use that for bootstraping and see what happens.
Come on now.
SAME IN NOSTR. My point isn't that DHT is BAD. My point is NOSTR IS NOT WORSE. Hear my point, not the adjacent one you think I'm arguing.
Nostr is clearly worse.
In specific ways yes, but specific claims about it should be narrowly focused on what is wrong, not overarching, because much of nostr is also better.
I know exactly what you are saying. But I don't care about Mainline vs Nostr. I only care about the technical details of Mainline because DHTs have too much bad reputation and I guess I am a bit autistic about it. You said that a certain 3 nodes are necessary, they are NOT, and there aren't any specific ones that are necessary.
This is strictly better and different in kind from mere redundancy of servers. It is a self healing mesh. That is not special though, Bitcoin p2p gossip and every overlay network is this way.
Ok. Thanks. I didn't mean "a certain 3" but I see how it could be taken that way.
But again same in nostr, you don't need to use purplepag.es to get started. From any relay you can discover many others and then find people's relay lists scattered about. It is a self-healing mesh too. Uglier and less certain, but not on a different order of censorability. Unless you are using primal (so I hear, I don't use it).
Yes everything can be anything if you try hard enough. You can even search Google to find the newest incarnation of your favorite pirating index after the previous one gets taken down. Hell you can download HOST.txt from your friends by sms.
But at some level of generosity technical discussions become too unserious
Not only is Nostr relays are orders of magnitudes fewer, and less time tested. and dns dependent. They are also not cheap to run AND have a conflict of interest since they double as service providers
This is the Mastodon and even Email fault, where the service providers can boycott and censor other providers.
But Mainline is cheap, extremely redundant and have no interest whatsoever in what is published on it. They don't compete with the services that are published on it.
This is significant in my opinion, you can't dismiss all of this merely as difference of degree... at some point it becomes difference in kind.
could nostr clients use pkdns for dns lookups?
I agree with that first stuff: fewer, less time tested, dns dependent (you can specify IP addresses though). And even worse you can only have one endpoint, and you can't change it without losing the relay's reputation.
I think they are cheap to run though. And don't have to also be service providers. I wrote chorus to be a personal relay so everybody could be on their own personal relay.
Sounds like you are arguing for decoupling resolution from service providing, and providing resolution where it can't be boycotted/censored. Sure you did that. But at some point there has to be a service provider so the problem is just conveniently pushed over to them.
You can change providers in Nostr and Email both, but not Mastodon.
When you change providers in Pubky you can hotswap without losing context thanks to pkarr.
Yes. Great stuff.
And in nostr when you change relays in your relay list, and move your data, you have done a similar thing. That is, USERS can move whenever they want, to evade censorship, etc. But RELAYS cannot. With pubkey in a sense, the RELAY is the USER.
But in nostr you can have one set of relays for sending things to the world (outbox) another set for receiving messages (inbox). In pubky you just have one, right? So on the data side it is less redundant? I could be wrong on that.
If I run my own Outbox, how can I tell the whole world about it if they 10 relays every one reads from censor me by decree from USG under the threat of unpersoning the relays owners aka sanctions?
If people only read from 10 relays under USG control, nostr won't survive that. I have 3,881 relays in my gossip client relay list. Granted most of those are junk. But there are far more than 10 relays. When I advertise my relay list I generally hit about 100-200 relays with it. For sure some of those are in China, Russia, Iran, etc. So this situation isn't close or likely.
This might be fine if nostr stays small for pocket/private networks, but for social media, marketplaces, etc... nope.
Good to hear, but this sounds like an adhoc gossip network. We don't need to reinvent overlay networks, we know that DHTs scale, gossip doesn't. Gossip only work when you are already cursed with the need to replicate a dataset in full.. for everything else, sharding aka DHTs are the optimal solution.
The 200 relays you reach, are not likely to be the ones I read from, and if it is likely, then it means these relays are under too much stress and they will churn or drop your data soon.
Nostr is good enough for sure, but physics is physics, and the Web needs a DHT or ICANN... these are the choices.
I agree DHTs are the superior solution. The odds that the 200 relays I used do not overlap at all with the ones you search are vanishingly low when we are inspired to use the most popular of relays. Is that a centralizing force? Yes, but a modest one, and what is popular changes ovrer time. Nonetheless it is clearly inferior to the DHT. Which is just one of the (many) reasons I'm working on Mosaic.
I honestly cant follow this degree of equivocation.
It just sounds like nothing means anything and everything is the same even though all of it is totally different...
Pubky currently supports modular optional indexer(s) and homeserver(s).
Indexer can pull from one or many homeservers, or anywhere really.
Apps can rely on indexer(s) or not.
I'm sorry that it sounds like meaningless equivocation to you. Maybe you don't understand enough about how nostr works.
In nostr, to find somebody you look for their kind-10002 relay list. The spec says to spread that thing as far as you can. Relays are all supposed to host it if they support NIP-65. Just one per pubkey, a newer one replaces the older one. Inside that signed document you specify which relays you are using. At any time, you can change those relays and publish a new relay list document.
If you are using 5 relays (3 for outbox, 2 for inbox), then one of those outboxes decides to start charging, you go find a new provider, copy your data from the bad relay (or better yet one of your other trusted outbox relays) to the new provider, and update your relay list.
Instead of registering an endpoint in the DHT, and updating that whenever you like, you publish multiple endpoints in a relay list and update that.
Outboxes are where people come to read your posts. Inboxes is where people drop messages to you. Because I don't follow everybody who replies to me, I can't expect to see their replies on their outboxes, so they have to copy me on my inbox.
So in nostr you specify MULTIPLE relays. I'm not sure but with pkarr you specify one homeserver? Maybe you can specify many, I don't know.
BTW having multiple relays has problems. People go off and setup hundreds and then our clients get overwhelmed. I think the spec should say "only first 3 count".
I am arguing for the decoupling, but not as a novelty, we have a history of Internet and web architecture to support this decoupling as an important principle. Nostr ignored that because it was designed to serve one purpose (social media) where highly contextualised posts are the norm.
Of course now the revision of "nostr is for the other stuff and fix the web" is here because the social media usecase stalled, but here is where i come to be annoying and say no stop, the mess that was made while assuming social media, doesn't fit when you are trying to fix the web.
If you want to fix the web, you should fix DNS, because that is the only minimal layer that actually can be decentralised at scale, and is sufficient for all small world applications.
Anyways I don't think we have disagreements. I just wanted to defend DHTs honor from even more misunderstandings, not by you, but by people who would have read your posts and took them at face value.
I'm sorry my post looked like a shit take on DHTs. I was really kinda pissed at the many claims I was reading from John that people are retards and nostr is highly censorable whereas pubky is perfect and uncensorable. So in that shitty mode with a desire to knock him down a peg I tried to make a subtle point that on first blush sounds entirely like I was making a different point. And that happens when I post hastily.
Understandable, but it was still your attempt to correct me, and I don't think you really did in the end.
A lot of people were extremely retarded this week, nostr is highly censorable at scale (which was my actual claim in most instances i think), and it can still be argued that pubky's design could overall fix the web.
Maybe there are other claims i made you didnt address, and i know my style is difficult, but my behavior and claims were all sincere.
Replace it with an empty list and see what happens.
Those three (router.bittorrent.com, dht.transmissionbt.com, dht.libtorrent.org) are convenience helpers baked into popular clients.
The DHT spec allows any UDP host to act as a router node, and they already do (uTorrent, qBittorrent, Aria2, PicoTorrent, etc.)
does pubkey have a baked in router helper?
Uhh yeah this is my question too
To be clear, you can use ANY NODE and it will work. And there are gazillions of nodes. But you still need to get started somehow. I don't think anybody could design a better system. But yes it has baked in starting points. see DEFAULT_BOOTSTRAP_NODES
https://docs.rs/mainline/latest/src/mainline/rpc.rs.html
My point was only that this is censorable, that EVERYTHING is censorable in this sense, and that nostr is just as censorship-resistant in this sense. Not trying to say the DHT isn't good.... it is great.
My point was that if you don't have a node list, you can't get one without going to these "convenience nodes". Take down these nodes and new people who have no other way of finding a node can't get anywhere.
This is IMHO the same problem nostr has, except nostr has the problem right now because it has no "convenience nodes".