Everything will be on relays. It's just a matter of time and enough scale.
Discussion
Hopefully not nostr relays which are not designed for binary data.
Specialized relays will appear. But in the meantime, everything goes into the regular relays.
Regarding text files, which is 99% of git repositories: Has there been any discussion about tokenizing the files and then storing the token “chains” separately from the token values? It would help with compression and redundancy. Might be shit for performance though.
I don’t agree with this at all. It’s way better to just put file metadata on relays and have file data hosted elsewhere. Not only for performance/scalability reasons, but also to minimize legal liability for relay operators.
Well, that's the beauty of nostr. Anyone can start doing it and there is very little we can do to stop it.
Indeed :)
After sleeping on it I’ve kinda changed my mind. I’m now favoring #[2]‘s approach. It gives accessibility options and if the file is changed or goes missing (404) then the relays can be queried for another URL for the same file. That’s kinda major. That to me is backwards compatibility and worth the hassle of the change.
Still, if we have the file data hosted elsewhere, shouldn't we at least have integrity hashes embedded within the metadata of kind:1 note?
Not funny if everyone's past notes start showing porn.