Magnus Thank you for the explanation: it does make sense indeed and I learned something today.
Up to the point of wondering whether having this 'no UDP dependency' written somewhere as 'non-invariant'. Regards -éric -----Original Message----- From: Magnus Westerlund <[email protected]> Date: Thursday, 7 January 2021 at 09:50 To: "[email protected]" <[email protected]>, Eric Vyncke <[email protected]> Cc: "[email protected]" <[email protected]>, "[email protected]" <[email protected]>, "[email protected]" <[email protected]>, "[email protected]" <[email protected]>, "[email protected]" <[email protected]> Subject: Re: Éric Vyncke's No Objection on draft-ietf-quic-invariants-12: (with COMMENT) Hi, One comment from my perspective here. On Wed, 2021-01-06 at 04:09 +0000, Lucas Pardue wrote: > > > ---------------------------------------------------------------------- > > COMMENT: > > ---------------------------------------------------------------------- > > > > > > == COMMENTS == > > > > Should the use of UDP transport be also an invariant ? I think that would be a very bad move. QUIC is designed for usage over UDP, it is not strictly dependent on it. In the last call there was some discussion to go the other way of rather being explicit about the requirements QUIC have on lower layers. That was considered a change that was to substantial to be done in the late stages. However, QUIC from my perspective have very little dependency on lower layer(s) as it provides datagram delimiter, a basic path identification (UDP+IP), and service demultiplexing based on server side destination port(s). Thus, I don't see that UDP would be an invariant, and from one perspective IETF is already designing a protocol in MASQUE that defines how QUIC over QUIC can be done, where UDP is only emulated on part of the path. Cheers Magnus
