[e2e] Can feedback be generated more fast in ECN?

[Michael Welzl]

> Let's see, SQ lets the router tell the end-point that the 
> network is busy,
> which lets the end-point deal with it by dropping cwin immediately.
> 
> ECN tries to do the same basic thing by building a bunch of 
> smarts in the
> middle of the network that may eventually after a while if congestion
> isn't complete link failure get around to telling the sender about it,
> after the sender has drained what's left of rwin, assuming 
> it's ECN-aware
> and that it doesn't take the ACKs as a cue to slide the 
> window over and
> send some more data for the network to think about.
> 
> Which is closer to the dumb-net/smart-node principle?

ECN.

If you want to do SQ, you need to be aware of the level of
congestion, too - be it based on the instantaneous queue length,
an EWMA of the queue length as in RED, whatever. So SQ and ECN
are similar in this respect.

When it all comes down to what a router has to do once it knows
about congestion, ECN merely has a router check a bit and possibly
set another one whereas SQ means generating a whole new ICMP packet.

I still don't think SQ is completely useless. But I agree with
Sally that it makes most sense for special cases (like TCP
over Sat) and that these special cases may require special mechanisms.

Cheers,
Michael