[quicwg/base-drafts] Confusion about number of outstanding CIDs (#1464)

[Mike Bishop <[e-mail]>]

Suppose an endpoint is probing an alternate path to keep it available for future migration.  (Probing periodically over cellular in case Wi-Fi goes away, for example.)  The endpoint has N connection IDs outstanding and chooses to use one of them on a probing packet, which gets lost.  When the endpoint retransmits the probe, it uses a different CID because it has been quiescent for a while.  This probe gets through.

At this point, the endpoint believes it has N-2 remaining CIDs to work with, while its peer believes it has N-1 CIDs still issued.  Now the text about making sure your peer has outstanding CIDs becomes pointless, because the peers can't accurately estimate.  Repeat this a couple times, and you might find yourself stuck behind "MUST NOT send additional packets until a NEW_CONNECTION_ID frame is received" because you've run out and the peer thinks you haven't.

ACK tracking would let the endpoint detect that this has happened (all packets using CID A have been declared lost, so peer doesn't know I burned it), but that doesn't permit the endpoint to *fix* the problem.  I think this ultimately leads us to a NEED_MORE_CIDS frame, unless anyone else has an idea.  (Not quite CID_BLOCKED, because once you're blocked it's too late.)

-- 
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/1464