Re: [quicwg/base-drafts] Clarify text around preferred address (#3589)

[Martin Thomson <[e-mail]>]

@martinthomson commented on this pull request.



> @@ -1084,10 +1084,10 @@ Sending a RETIRE_CONNECTION_ID frame indicates that the connection ID will not
 be used again and requests that the peer replace it with a new connection ID
 using a NEW_CONNECTION_ID frame.
 
-As discussed in {{migration-linkability}}, endpoints limit the use of a
-connection ID to a single network path where possible. Endpoints SHOULD retire
-connection IDs when no longer actively using the network path on which the
-connection ID was used.
+As discussed in {{migration-linkability}}, endpoints MUST limit the use of a
+connection ID to packets sent from a single local address.  Endpoints SHOULD
+retire connection IDs when they are no longer actively using the network path on
+which the connection ID was used.

Perhaps the answer here is to say "endpoints limit the use of a connection ID to a single network path".  The point about NAT rebinding/inadvertent migration here is not relevant and somewhat distracting.  The reintroduction of the "MUST" is problematic as this implies a canonical requirement, but you can't really make that without adding the necessary context about NAT rebinding.

> @@ -2277,11 +2278,21 @@ linked by any other entity.
 At any time, endpoints MAY change the Destination Connection ID they send to a
 value that has not been used on another path.
 
-An endpoint MUST use a new connection ID if it initiates connection migration as
-described in {{initiating-migration}} or probes a new network path as described
-in {{probing}}.  An endpoint MUST use a new connection ID in response to a
-change in the address of a peer if the packet with the new peer address uses an
-active connection ID that has not been previously used by the peer.
+An endpoint MUST NOT reuse a connection ID when sending from more than one local
+address, for example when initiating connection migration as described in
+{{initiating-migration}} or when probing a new network path as described in
+{{probing}}.
+
+Similarly, an endpoint MUST NOT reuse a connection ID when sending to more than
+one destination address.  However, if an endpoint receives packets from a new
+source address with the same destination connection ID, it MAY continue to use
+the current connection ID with the new address.

How about:

> However, if a peer experiences NAT rebinding, an endpoint could receive packets from its peer using a new source address with the same destination connection ID.  If a peer is first to use a connection ID on a new network path and that connection ID was originally used on another network path, an endpoint MAY also use a connection ID that it used on the original network path.

That is much more wordy, but it might be clearer.  The point is to say that you don't reuse a connection ID unless your peer reuses first.

There is a wrinkle here in the case that an endpoint has multiple valid addresses (think preferred address).  In that case, we don't want to allow reuse as that implies a bunch of other things.  The fix there might be to add:

> This exception that allows the use of a connection ID on multiple network paths only applies when the remote address changes and the local address remains constant.

-- 
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/3589#discussion_r422718955