[e2e] ECN vs SQ: reverse path congestion

[Ratul Mahajan]

Reading the thread on ECN vs SQ has made me wonder how does ECN compare to
SQ when there is congestion on the reverse path. Apart from the 
comparison, what should the right response be in either case?

Do ECN specs state that TCP sender should cut down when it receives CE bit
set in the ack packet (assuming acks are declared ECN aware; i don't know
this detail either)? Should the sender cut down at all? It seems yes,
because acks are generated by the data packets only. (Or if we have
smarter receivers, they could generate less feedback when the reverse path
is congested; who tells it that?).

Assume we had SQ in place and the receiver gets an SQ message. What is the
right response then for the receiver? 
i) generate less feedback, which may be ok for protocols with cumulative
acknowledgments (TCP). Taking this to the extreme, the sender would
eventually have to cut down.
ii) explicitly inform the sender to cut down.

Based on the responses, which mechanism is better at dealing with reverse
path congestion?


Cheers,

	- Ratul