The MPLS WG Archive

[Ping Pan <pingpan@cs.columbia.edu>]

Curtis Villamizar wrote:
> 
> In message <5.0.0.25.2.20000929150628.0274b540@flipper>, Fred Baker writes:
> > At 01:08 PM 9/28/00 -0700, Randy Bush wrote:
> > >this has appeal.  but ecn looks much simpler.
> >
> > Is ECN something that you would consider pushing toward standardization? I
> > like it, but I'm looking for operator feedback, and some (notably Juha)
> > distrust a mechanism that trusts the host.
> 
> It should at least go through as experimental but preferably as
> proposed standard.  Shouldn't the ECN WG get Cc'd?
> 
> Curtis
> 
> ps - Juha doesn't have to turn it on if he doesn't trust it.

Please educate me here. Is the idea of ENC similar to that of FECN and
BECN in Frame Relay? That is, upon congestion notification, end nodes
start to slow down packet transmission. Many routers today happen to be
the "end nodes". They generally use GRE encapsulation to support user's
traffic (such as delay-sensitive SNA data). Supporting ECN would mean
that the edge routers must (re)implement congestion avoidance mechanism
for encapsulated packets, is that correct? Do you think this may cause
deployment problem? Or maybe all routers have such capability already.

--
Ping Pan         http://www.cs.columbia.edu/~pingpan