Also "just follow curated, moderated relays" = Mastodon. Change my mind!
Discussion
The data doesn’t live on just one relay, and npub doesn’t belong to any relay either, so it’s not the same as Mastodon.
I guess what fiatjaf wants are relays that curate high-quality content, each with their own filtering strategies to provide feeds tailored to different interests.
Outside of npub portability, again, it seems like you're just describing Mastodon. Federation means different instances can indeed share information, and I guess there is technically a means of user migration between instances, even if it isn't frequently used... But in what fiatjaf is describing, you wouldn't be moving your npub at all anyway!
"Just connect to these three or four servers" just seems to defeat the entire purpose of a decentralized social network. And besides that, the issue is that it vastly increased the legal liability on relay operators, who now 100% have to operate as content moderators.
Strictly speaking, there's no such thing as "migrating" an npub, it was never tied to any one place to begin with.
Once an event is signed, it objectively exists. In fact, your events already exist on relays all over the world, even ones you never explicitly published to. (This very reply of yours is already on a relay I run at home.) That’s what I see as the biggest difference between nostr and mastodon.
Decentralization isn’t about how many relays you use. You connected to 14 relays, but I’d guess you don’t have absolute control over any of them. I only use five, but I control two of them. So who’s more censorship-resistant?
Any service that exposes data publicly carries some legal risk, that’s a general truth, not specific to this idea. And no one’s saying every relay needs to moderate content. Public relays and curated relays can coexist.
Or maybe self-signed individual json events are enough. Trying to create a quantum superposition of yes global state + no global state is goofballs, has always been. This is a much better idea.
this concept of self-sovereign identity is not common on the internet. it's used with SSH, and a small amount of TLS works from this basis. the DNS roots are sovereign in as far as a small number of individuals have the secret to sign stuff, and this is why it's robust.
the rest of the internet tho. they don't care. in my fiat mine job, we are working with several blockchains and fortunately two of them that we work with the users' signatures are generated locally by the users, albeit with the help of this "web3auth" system which binds access to a secret key to an email address. we are integrating another chain, an ethereium fork sponsored by Sony, and the web3 devs in charge of that yet again chose a non-self-sovereign identity scheme whereby the secret bound to a user's email related identity depends on a smart contract to ... idk even how this is considered secure... but the smart contract signs their events.
i haven't implemented the prescribed API call to that SC to verify the signatures, and as such, an important but minor attack vector on our users is currently open because the server i built isn't validating the signatures. but this is a bad thing. and its a very bad design to shift authentication inwards to the centre of a network system. authority to sign events should be on the edge, this is what "self sovereign identity" means.
the history of hacks on central authentication systems is extensive. why people keep building them is beyond me.
Agree. We were trying to think of a term for it for some of our internal stuff, "end of the road authority" is our placeholder, but "authority on the edge" also works. That alone, if well done, is something.
Circling back, many of Nostr's problems come from trying to have global-state features without paying the global-state price and then getting slapped around by the debt collector (network physics).
The clobbering of replaceable events like follow lists is the perfect illustration of this. Network physics is like "Did you pay the price for this little global state you're trying to sneak in here?" and the kind3 event spec is like "umm.. no..." and then it's the baseball bat again.
I would prefer a p2p nostr over a multi-server model
The definition of Nostr is the multi-server model.
I don't think there shouldn't be servers, I just think every person should be running one. the community relays should also exist, but relays count should > npub count since every npub has a relay.
those personal relays might act like aggregators for whatever that person is interested in, so the "50 people" you follow at their personal relay gets you content from many more sources.