Re: [quicwg/base-drafts] Reserve Long Header Packet Type for QUIC-LB (#1980)

[Martin Thomson <[e-mail]>]

I think that we probably want to have a deeper conversation about this draft.  I look at it and ask why you need to create a bespoke protocol here at all.  If it is going to look like QUIC proper, why not make it be QUIC proper?

But even assuming that you do need that, why are you concerned about making this look like QUIC proper?  I can't see why you would need that for routing purposes, because it seems like you should be back to 5-tuples for routing on this bit.  I can't see why you would concern yourself with ossification on a link that you presumably have considerable influence over.  I can see why you might want to disambiguate this connection from "real" connections, but you could use a separate port, ALPN, or even a new QUIC version if it came to that (though I would advise against a version).

On response to @marten-seemann's comment, which I agree with, you might also like to consider the effect of having 5 types in the long header and what that does for protocol ossification.

-- 
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/1980#issuecomment-437724235