Re: [quicwg/base-drafts] Connection ID Length changes (#2473)

[David Schinazi <[e-mail]>]

I think banning switching from non-zero-length CID to zero-length CID might be too constraining. Imagine a server is initially reached via an IPv6 anycast address, and it tells the client its IPv6 preferred address. If that server owns a /64, it can give a different preferred address to each client which allows zero-length CIDs and migration without requiring trial decryption because the server can find the connection simply by looking at the packet destination IP address. (Note that kernels don't let you easily do that today but that can be fixed.) The only downside I'm seeing here is that migration can now be tracked across paths since the connection ID doesn't change and the server IP is enough to identify a connection. I suspect there are deployment scenarios where reducing CID overhead is more important that the privacy risk. But, given that the server preferred address is only advisory, a client that cares more about privacy than about the CID overhead can simply refuse to switch over. With all that said, I don't think the restriction against zero-length-CID helps us.

-- 
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/2473#issuecomment-466140260