What is the crowd's opinion of Hypercore Protocol?

From 3.5 billion comments

Ti permette di creare un sito web e di fare hosting personalmente. Poi questo sito puoi condividerlo tramite un url generato dal browser grazie ad un protocollo chiamato Hypercore.

Gli url sono del tipo hyper://qualcosa.

Il resto funziona simile a bit-torrent solo che invece che fare seed di files, fai hosting dei siti. Puoi decidere di hostare anche siti di altri peer per mantenerli online.

È tutto distribuito e non ha quindi bisogno di un server, un sito è raggiungibile finché c'è almeno un nodo acceso.

Il browser supporta anche il classico protocollo HTTP, quindi è possibile anche usarlo come browser generico. Ma diciamo, ce ne sono di migliori.

Yes! This is common in distributed systems, like the InterPlanetary File System or Hypercore

The concept is the same: You have a protocol, and a location of a resource that can be acquired using that protocol, joined together like ipfs://QmZtmD2qt6fJot32nabSP3CUjicnypEBz7bHVDhPQt9aAy

There are some similarities with Criptext but fundamentally there are two very different projects. Criptext is built on top of the Signal Protocol, Telios is built on top of the Hypercore Protocol.

I would say two of Telios' primary differentiators are that you own 100% of your data as it is stored on your local devices (with the possibility of server backup) and it is peer-to-peer service.

Our long term goal is also to provide users the ability to run their own server nodes on the network, further enabling users' ownership and control of their email.

>free services

We have a lot of 'free' services, but 'rational agents' would also be advertisement-averse. But otherwise I fully agree. The network effect of popular platforms is a giant hurdle for any alternative platform, especially if the benefits come in form of privacy and control. Though people do seem to have appreciate some kind of control, for instance choosing to keep the dislike count!

>p2p

P2P doesn't just mean crypto, which more naturally lends itself towards money and markets. It can also be oriented towards information sharing, akin to Bittorrent. Check out https://hypercore-protocol.org/ and https://cabal.chat/ for instance. Chats, social media and cloud services without servers or the blockchain.

>language models

I study language modeling myself, though mostly from a bit more traditional statistical perspective. To me data concentration is the biggest problem, all SOTA models are usually published so you can replicate the code at least. Github / Microsoft can train on all code that they host, FB / meta has uniquely large datasets of personal chat logs, etc. It does sound kinda scary to have them overfit a huge transformer model so it actually spits out someones personal information verbatim. Or just the ability to have a monopoly on the best language model for a wide array of tasks.

Thanks for all of your great questions!
 
We're using Hypercore protocol which uses something called Hyperswarm to connect peers over DHT/Kademlia. What's really neat about a Hyperswarm version that was released is you can create a firewall and whitelist specific public keys of peers that are allowed to connect to you. If a peer is not whitelisted then no exchanging of IPs happens.
 
We've also thought about eventually adding a mix net like I2P over the network for added anonymity.

We can treat files like torrents, but they aren't right now. We aren't using torrent tech, but something called Hypercore Protocol. It's works a little like Git + Torrent, which gives us the ability to version files and for them to be mutable which torrents do not allow.
Something that makes this service unique apart from other p2p-like services is that we can make connections to other peers that are behind firewalls and on cellular networks without needing servers to proxy those connections.

In a sense. I didn't change the wording from the site, but I guess the real point here is that the two ends find eachother via the hypercore protocol, which is itself p2p/distributed, and not DNS or something. DNS is arguably distributed but very hierarchical and centralised really - all authority ultimately lies with the 13 dns root servers. At least if you don't use a dns alt root of course - but alt roots still have all the weakness of the DNS system plus of course less adoption.