Oddbean

Exactly, so using relay hints that list relay URLs in an immutable note is stupid — if the relays go down, the post is toast. Gossip is a step in the right direction, but it isn’t flawless either due to requiring relays to constantly spread your gossip note. 🗒️

Not to mention there is not a standard way for devs to implement it. You can implement it wrong or “right”? The idea that we are getting upset at devs for not implementing nostr right is silly. What is the right way? No one has ever done this write up. How about write a tutorial or nip? This is much more productive than just shitting on devs for doing the simplest thing when they first get started (relay pool). I guess @fiatjaf is calling me and many others for not implementing “fiatjafs vision”, but what is this vision exactly? following relay hints? this doesn’t solve everything, the relay hint in a note could be down. nevent relay hints? nevent didn’t even exist when I started building a client. How is this “obvious”? How do you select the ideal relays for many different types of queries without blowing the connection budget? What client is @fiatjaf building that has this perfect algorithm? Please teach us.

Honestly jb, I’m going to declare independence from fiatjaf’s madness soon. I’ve been interested in @elsat nostrability for maintaining compatibility across clients, as I’m sure you’re aware. I’m working on a clear standard (list of cleaned up NIPs etc) that others can follow if they don’t want to try to sift through fiatjaf’s madness. It will avoid relay hints for discovering missing content, and other bad/messy ideas (like NIP04) — distilling the best of #nostr. Most of it aligns with Damus etc, but I’ve got a new outbox method that I’m eager to share soon. It doesn’t rely on the blockchain at all… it’s free and fast. You all can be the judge of it soon. Nostr is not what fiatjaf says it is. He helped start it, but I suspect the people are going to be the ones to scale it. People like us. Personally, I see fiatjaf almost always arguing in a disingenuous way — then he blames others for not knowing what to do. As long as we create a standard that many clients can follow in a compatible way, we don’t need his mad king behavior.

An old spec of mine tried to solve the consistency problem of Nostr by putting merkle roots of profiles on-chain. It was just too expensive for users & totally not scalable. Back when I first joined Nostr two years ago. I can admit when I was wrong. Have iterated since then.

nostr:nprofile1qqsrhuxx8l9ex335q7he0f09aej04zpazpl0ne2cgukyawd24mayt8gprfmhxue69uhhq7tjv9kkjepwve5kzar2v9nzucm0d5hszxmhwden5te0wfjkccte9emk2um5v4exucn5vvhxxmmd9uq3xamnwvaz7tmhda6zuat50phjummwv5hsx7c9z9 could provide automated Tests as requirements so people who implement it can be sure it's working right

Even with perfect Nostr implementation, frequent downtime of smaller relays leads to empty profiles and threads, highlighting the need for more robust and reliable relay infrastructure. 🎯🙌🫡