Re: [quicwg/base-drafts] Make transport parameter ID and length varint (#3294)

[Marten Seemann <[e-mail]>]

In addition to the points that @nibanks mentioned, there's another advantage of defining our own format for transport parameters: 

- We can require the list to be ordered by parameter ID (I've suggested this before somewhere, but #3169 doesn't yet incorporate this suggestion).

This would make the detection of duplicate parameters (which is now just a SHOULD due to the computational complexity) trivial, and would allow us to turn the SHOULD into a MUST. Furthermore, it reduces the profiling surface exposed by a QUIC implementation: I would assume that today it's probably possible to identify a QUIC implementation just by looking at the order of the transport parameters it sends.

@kazuho

> I see that some people are arguing that the encoding can be different based on the view that TP belongs to QUIC. I dispute that. We deliberately decided to mix crypto handshake and transport negotiation in QUIC. TP is a cross-layer thing that is exchanged during the TLS handshake.

I disagree with this statement. The reason we're putting transport parameters into the TLS handshake is because they're sent in Initial and that's basically the only way we can detect a modification by an on-path attacker. By putting them into the TLS handshake they become part of the transcript and therefore prevents this attack.
If there was a way to send transport parameters in encrypted packets, there would be no reason not to define a QUIC SETTINGS frame for them.

-- 
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/issues/3294#issuecomment-564844048