Re: [quicwg/base-drafts] Rework Key Update (#2237)

[Kazuho Oku <[e-mail]>]

@marten-seemann We cannot enforce implementors to code correctly.

The issue with the 2-bit KEY_PHASE design is that the receiver has a code path that is very rarely being executed (retaining 3 keys at once). Some stacks might fail to implement it correctly. Or some stacks might not implement key updates correctly at all. Anyways, we might see interoperability issues in the wild, and the bug would be hard to identify, because it's rare and because it would look like an connection stalling to the end user.

"MAY_UPDATE" bit takes a different approach. We do not have a code path that is rarely taken on the receiver side. Hence less chance of having interoperability issues. A receiver that does not allow the sender to update keys can be detected, and then the sender can close the connection with a specific error code. That is much better than the connection silently breaking.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/pull/2237#issuecomment-450019472