search

quicwg / multipath

In-progress version of draft-ietf-quic-multipath
Other
49 stars 17 forks source link

Should path status sequence be global, or per frame? #280

Closed huitema closed 10 months ago

huitema commented 10 months ago

I think we can use the definition that we have, but here is a corner case:

1) Peer sends PATH_STANDBY, path 1, sequence number 1 2) The packet containing that is delayed 3) Peer sends PATH_AVAILABLE, path 2, sequence number 2 4) The packet carrying PATH_AVAILABLE, path 2, sequence number 2 is received. 5) The delayed packet carrying PATH_STANDBY, path 1, sequence number 1 is finally received.

With the current spec, the peer is supposed to ignore the PATH_STANDBY frame, because it has received another status frame with a higher sequence number.

The global status number is fine, but I would prefer to test it by path, rather than globally.

huitema commented 10 months ago

Oops. This is fixed already, text is fine.