Oddbean new post about login | logout Unsubscribing from discussions about NIP 96. It's getting increasingly chaotic, and I think blossom is much better. If not because of the technical details, then because it's being built by people who really understand the strengths of nostr's architecture.
what do the NIP-96 people get wrong?
Nothing, the blossom people just get it really right.
How does one "subscribe to a particular NIP discussion" ?
on github ^^
I haven’t made a deep dive in NIP-96 but Blossom does make sense in many ways it’s worth to try and adopt it. So yes I’m a fan of Blossom. It’s early but I’m convinced it will evolve in a solid solution for blob distribution via Nostr.
Im a fan of blossom too. Though conent addressable files on servers is not a new idea. One slight issue is that it pollutes the name space, because not every storage provider wants to, or can, put files in the root of public_html. I dont think it's a good idea to distribute blobs via relays, because they are too big. Better to use references, imho.
In the current spec those blobs are not stored on relays perse right? I'm running a Blossom mediaserver instance which is not a Nostr relay. Just to be clear, with distribution I didn't meant the blobs has to be stored on relays. I agree that's a bad idea for many reasons. With the distribution I mean the the transmitting / advertising part of Nostr events (kind 10063) referencing the servers where your blobs are stored :)
Gotcha. Yes that's fine. I'd possibly think of that as discovery. Sorry for the confusion. All good.
nostr:nevent1qyw8wumn8ghj76r0v3kxymmy9e3k7unpvdkx2tn5dahkcue0qy2hwumn8ghj7un9d3shjtnyv9kh2uewd9hj7qg6waehxw309ac8junpd45kgtnxd9shg6npvchxxmmd9uqsuamnwvaz7tmwdaejumr0dshsqgzyas0qh2cms3gkrjnlfh4hxr8sad7vdl52a49pzxjtgk3rr0p265k3chld
Exactly… NIP96 always uses the original file’s hash in the url even when the content was transformed, I.e. then the content returned does not match the hash in the url. Mirroring/Reupload with the same hash is therefore impossible. When transforms are disabled (there is an option) then NIP96 behaves very similar to blossom in terms of hash usage (except for paths and auth events). Kieran nudged me to implement both on bouquet.slidestr.net and they play along nicely when not using server side transformations.
I like noStrudel too… but this is an interesting point about interoperability. I’ve been trying to parse that out here. Thanks for your contribution to my thought space.