Re: [quicwg/base-drafts] Don't store or retransmit PATH_RESPONSE frames, avoid buffering (#2729)

[Eric Kinnear <[e-mail]>]

erickinnear commented on this pull request.



> @@ -1721,7 +1721,12 @@ it can associate the peer's response with the corresponding PATH_CHALLENGE.
 ## Path Validation Responses
 
 On receiving a PATH_CHALLENGE frame, an endpoint MUST respond immediately by
-echoing the data contained in the PATH_CHALLENGE frame in a PATH_RESPONSE frame.
+echoing the data contained in the PATH_CHALLENGE frame in a PATH_RESPONSE frame,
+unless it has PATH_RESPONSE frames buffered for the same destination connection
+ID and wishes to limit memory consumption.
+
+An endpoint MUST NOT store or retransmit a PATH_RESPONSE frame, as a new
+PATH_CHALLENGE frame will be sent if another PATH_RESPONSE frame is needed.
 

I think the original intent here for “store” was in terms of (frames in) packets that are waiting to be acked, not things that might be kept around for a very short time to satisfy pacing and/or congestion limits on their way onto the network. Some of that should be cleaned up soon.

You should send multiple path responses in response to multiple path challenges, even on the same connection ID - unless you’re getting attacked with them at such a high rate that you can’t get them out the door quickly enough and you need to limit your memory consumption while doing 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/2729#discussion_r288411344