Re: [quicwg/base-drafts] Normative Text in Invariants (#3773)

[Martin Thomson <[e-mail]>]

I agree that this is editorial, but I think that we should say something that is subtly different.

There are two parts to your PR.  One part that attempts to describe what is normative in the document.  The other part attempts to more clearly describe the intent of the document: basically that anyone who looks at QUIC from the outside can only see and use this much of the protocol.

For the "normative" piece, we can probably say something less concrete but more principled.  That is: a protocol that follows the requirements in this document may call itself QUIC.  And that any attempt to treat QUIC generically MUST be limited to using the characteristics that are included in this definition.

For the intent part, I would say that the goal is to carve out a space for endpoints to communicate without interference from others.  The goal is to design a protocol where participation by others only occurs with the express permission of endpoints.  (I think we should be careful not to say whether this is just one or both endpoints granting permission here, because QUICv1 is a little inconsistent on this point itself.)  That's somewhat stronger than the position you take.

Leaving "others" unspecified here is good.  I see that you have tried to define those others, but I don't think that's helpful.

The nice thing about this approach is that it requires far less text.  Your PR doubles the size of the draft.  

BTW, I do like the MUST regarding treatment of unknown versions.  That is, if an entity doesn't understand what is going on then it can only apply generic treatment.  But we all know that's RFC 6919 language.

-- 
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/3773#issuecomment-645709658