Hi MSavoritias, Le mercredi 3 juin 2026, 08:14:44 heure d’été d’Europe centrale MSavoritias via Standards a écrit :
I do not understand why the XEP is locked to OMEMO and also we roll our own identity scheme.
This protoXEP is not locked to OMEMO at all, why make you think that? OMEMO is only used for the optional automatic private key synchronization feature, and there is an alternative out-of-band QR code based method specified. This not really a scheme, it's inspired from existing solution (Peer ID from LibP2P), and it's simply using an existing widely used algorithm (Ed25519). This spec simply adapt it to XMPP mechanisms, a thing we would have to do for any solution.
What we can do instead is use a standard that is well documented and already exists for this Decentralized Identifiers
I admit that I was not aware of DIDs, but we could still have to adapt it to XMPP, and it's far more complicated to implement. The current proposal is simple and use an algorithm that most client already use. Also, the prefix byte is made to update to another method if necessary, so DIDs or anything else could be used in the future.
https://www.w3.org/TR/did-core/ this would give us:
- decoupling of the verification of identity from any centralized entity and any encryption scheme
This spec does that. The prefix byte is there to not be dependent on current encryption algorithm.
I understand that it supposedly strives to not depend on servers, but such a thing is not possible yet in xmpp due to DNS reliance. DIDs on the other hand can support multiple verification methods.
The main DNS reliance is in the JID. This proposal (again made to be easily mappable on Peer ID which is widely used) is made to not have to rely on DNS.
- our own xmpp method to resolve things to
- a standardized way to represent this and do discovery, not reinvent our own
- multiple ways of authentication besides just xmpp pubsub things
This specification use PEP which is also what is used for OMEMO and OX, and many other things. It's not reinventing anything. And other authentication method are planned, notably out-of-band via QR Code. PEP doesn't have to be the only way, it's just convenient because it has everything needed. Again, this is a protoXEP which would go as an experimental specification, made to test implementation, and see if it works or not, and how it can be improved. We are not discussing here a move to draft or final. Best, Goffi