PSQv2 #1048
PSQv2 #1048
Conversation
jschneider-bensch
force-pushed
the
jonas/psq-v2
branch
from
9dd51a1 to
75e432e
Compare
|
I've now implemented registration mode as well, and addressed your comments. In registration mode, the inner initiator payload is passed out for an application to validate, as discussed. |
libcrux-psq/src/protocol/keys.rs
Outdated
| } | ||
|
|
||
| pub(crate) fn serialize_encrypt(&self, payload: &impl SerializeBytes, aad: &[u8]) -> Vec<u8> { | ||
| let mut ciphertext = vec![0u8; payload.tls_serialized_len() + 16]; |
There was a problem hiding this comment.
If we want to follow snow APIs, we should have the caller always pass in the output bytes to write into. That's usually nicer for networking to avoid coyping things around.
There was a problem hiding this comment.
So you mean the application ultimately will pass in output bytes?
| Message { | ||
| ephemeral_ecdh_pk: responder_ephemeral_ecdh_pk, | ||
| ciphertext, | ||
| aad: aad.to_vec(), |
There was a problem hiding this comment.
aad should maybe be owned here again.
There was a problem hiding this comment.
I think it's still not consistent what's owned and what not everywhere. I think I had some trouble with serialization when there were named lifetimes.
| pub(crate) key: AEADKey, | ||
| } | ||
|
|
||
| pub struct SessionState<'keys> { |
There was a problem hiding this comment.
How is this session state being loaded/stored? And how are the referenced keys being filled in when this happens.
There was a problem hiding this comment.
I haven't implemented session storage yet, but the idea was to use tls_codec for that as well. But indeed I wonder how the references should be handled
Fixes #1044
Fixes #1043