Scott Courtney follows up on his recent article about ECN with
some additional notes on how rapidly the protocol is gaining
acceptance in the real world, and special issues Linux advocates
face when reporting problems with non-compliant networks:
"My recent article "Why Can't My 2.4 Kernel See Some
Web Sites," generated some vigorous discussion in e-mail, and in
the LinuxToday talkbacks.
Several readers wrote to say that the ECN protocol is
technically still experimental, and that it therefore should be
expected not to work correctly with some servers. Others noted that
nearly all distributions using a 2.4.x kernel have ECN turned off
by default, and that you have to deliberately enable it in order to
encounter the problem. Finally, several readers pointed out that
the failure of an ECN-enabled client to connect to a server is a
problem at the server end, or at a router or firewall in between,
and not at the client itself. These folks mostly urged Linux users
who find ECN-related problems to report them to the site
administrator rather than just turning off ECN.
While the first article was technically correct, those who have
written in with comments have raised some very valid additional
points, and I would like to address those here. The commercial
distributions do in fact have ECN turned off by default, so many
users of 2.4.x kernels will never encounter problems. The kernel
help for this option reads as follows in 2.4.2:"