[e2e] Multiple TCP-friendly Sessions and Cong. Control inuser-mode?

[davide+e2e at cs.cmu.edu]

> Packet losses can be caused by events unrelated to network congestion,
> e.g. noisy wireless links, routing events, and faulty hardware.

I'm not sure there is consensus that well-designed wireless
links export end-to-end packet loss bursts, with the possible
exception of high-latency satellite links, which pose additional
challenges.

If an *occasional* routing "event" (due to a link going down) dumps
a bunch of packets belonging to a single connection, inferring
congestion could be both prudent and accurate.

If routing "events" *frequently* dump a bunch of packets belonging
to a single connection, something is broken.

The statement that loss does not indicate congestion for poorly-designed
wireless links, broken routing setups, and faulty hardware is probably
true, but hard to get fired up about.

Dave Eckhardt