On 14/10/2022 21:06, Cordeiro, Cliff wrote:
Hi,
The loss timer as defined in RFC 9002 A.8 has an exponential backoff
component “* (2 ^ pto_count)” that is unbounded. This means that if
there is a lengthy network outage, QUIC may not recover until up to
twice the length of the outage because it will wait for the timer to
expire and the timer doubles without bound. For example, a network
outage of one minute could translate into a QUIC “outage” of two (on
average) to three minutes (unlucky upper bound).
I think there should be a limit to the loss timer. I don’t know if the
spec should provide a specific limit but I think it should be
allowable to set an upper bound on the loss timer in the 8-64 second rage.
Thank you for your consideration.
Cliff Cordeiro
RFC 8961 provides Best Current Practice with high-level requirements for
time-based loss detectors appropriate for general use in unicast communication
across the Internet.
https://datatracker.ietf.org/doc/rfc8961/
It refers to RFC6298, and states:
A maximum value MAY be placed on the RTO. The maximum RTO MUST
NOT be less than 60 seconds (as specified in [RFC6298]).
Gorry
