Re: [quicwg/base-drafts] Request to Retire Locally Issued CIDs (#2769)

[MikkelFJ <[e-mail]>]

mikkelfj commented on this pull request.



> @@ -994,6 +994,14 @@ packets sent from only one local address.  An endpoint that migrates away from a
 local address SHOULD retire all connection IDs used on that address once it no
 longer plans to use that address.
 
+The endpoint can explicitly request its peer to retire connection IDs by sending
+a NEW_CONNECTION_ID frame that contains a Retire Prior To field.  The peer is
+required to retire these connection IDs in a timely manner.  On receipt of the
+acknowledgement for the packet that contained the NEW_CONNECTION_ID frame, the
+endpoint may start a 3 PTO timer, which on expiring may close the connection
+with a PROTOCOL_VIOLATION error if all the connection IDs were not retired as
+requested.
+

Yes, I was thinking 3 RTT. 3 PTO may be enough, but an attacker could still delay so I prefer to drop such packets.

If you remove them from the routing table that is a deployment strategy - but most would likely keep a grace period before resetting or dropping. I'd imagine the middlebox load balancer is slow to update and only drops while backends are more aggressive because they are directly targeted. This is actually also true without explicit retirement - do we have any text on how to deal with that?

-- 
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/pull/2769#discussion_r290513357