Re: [quicwg/base-drafts] Equivalence of preferred_address and NEW_CONNECTION_ID (#3560)

[ianswett <[e-mail]>]

@DavidSchinazi I agree with your intuition, but I think we discussed this in Zurich and ended up concluding the preferred_address CIDs weren't special.  This was in issue #3348 and also came up in #3353 (which I realize hasn't gone out for a consensus call yet, that PR seems stalled?).

In the upload example, the server needs to know that the client is going to send a lot of data, which it might know via SNI, or it might not.  I tend to think the more annoying restriction is that you can't change from a non-0 length CID to a 0-length one and vice versa.  By supplying a 0-length CID later in the connection along with other CIDs, the client could choose whether it wanted to use a 0-length CID at any time.

Thinking out loud, an easier solution may be a TP like 'accept_empty_connection_id' that indicates at anytime, the peer can switch to a 0-length CID if it wants.  It could either think the 5-tuple is stable or its happy re-establishing the path with a non 0-length CID after a NAT rebind.  In this world, the Connection ID is more of a path re-establishment token.

-- 
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/3560#issuecomment-611593332