Iβve been noodling on my OpenSats projects and one thing I wanted to hear peopleβs thoughts on is the idea of lightly encrypted groups vs. relay-based groups. And by lightly encrypted I mean that all group data is encrypted with a shared key that gets rotated, but without end-to-end encryption, forward secrecy, post-compromise security, and all the fancy stuff you get with MLS. Basically the unmerged NIP-87 (

GitHub
Add closed communities by staab Β· Pull Request #875 Β· nostr-protocol/nips
This is an attempt at superseding #706 and incorporating existing NIPs for community definitions and member lists. The design is worse in many ways...
)
I know this idea has been discussed a lot, and I have been pretty convinced that NIP-29 made the most sense for the most groups. I also know MLS groups are in the works, but they have a lot of downsides. So a few things over the past month are making me reconsider.
The main one was talking to
@n1.g1.n1 from
@npub1j4g0...fuu4 who makes a good argument that groups should be a first class citizen on Nostr. This would enable groups of groups and potentially other innovations like putting the group master key in a FROSTR cluster. It also helps enable forkable groups and groups migrating between relays / sets of governing rules. (Great article from SocialRoots about their full vision

Socialroots
Intimacy Gradients: The Key to Fixing Our Broken Social Media Landscape
Un-breaking our social networks requires fundamentally different architectures than the defaults currently used to build most tools for coordination.
)
Another factor is that people keep asking me if groups are going to be encrypted in my new client and I donβt like saying no to that π
. Even though I think the confidentiality guarantees of NIP-29 are good enough for most groups - thatβs not what people want to hear. I used to think that getting a bunch of Nostr clients to all implement key rotation the same way was too much to ask, and I still think MLS is overkill for medium to large groups. But if you allow some privileged software to run with some kind of group admin key to do the rotation (an allowance that NIP-29 already makes) then it hugely simplifies the complexity for client developers and now you can say the magic word β¨encryptionβ¨.
I also feel like I missed out a bit on the debate between these when it happened. What do you think?