[e2e] Interpretation of ECN as a less severe congestion signal

[Michael Welzl]

Hi all,

Just out of curiosity:

Why is it that the semantics of an ECN flag were not defined
as a LESS severe congestion signal than a dropped packet?
For example, a sender could reduce the rate by 1/4 instead
of 1/2 in response to an ECN flag; this way, ECN would provide
MORE instead of complementary information.

Router behaviour would then be:
minThresh <= queue length < maxThresh: mark packet
maxThresh <= queue length: drop packet

I assume that this can't be changed now because the behaviour
in routers needs to be uniform - but I like the idea ... it's
simple, and it should be better ... it's well known that the
rate reduction factor (beta) could be higher than 1/2 and the
AIMD idea would still be preserved.

I've seen this idea in the Multilevel ECN proposal by Arjan
Durresi, Jain et al (somewhere on Raj Jain's publications site).

Cheers,
Michael