[e2e] What should e2e protocols know about lower layers?

Christian Huitema huitema at windows.microsoft.com
Thu Oct 11 14:42:08 PDT 2001


> > It should be pretty obvious that suppressing congestion control
> based on a
> 
> > naïve test such as "same subnet mask" is BAD.
> 
> 
> If it's that obvious, why do we allow broadcast (and broadcast-based
> multicast) on that same test?

There are two issues with broadcast, congestion control and the number of systems that your broadcast rains on. The "same subnet" test has some value as an estimate of the number of systems. The worse case are switched local networks that could reach a few thousand stations; if I remember correctly, the old pathological example was CERN in Geneva with about 40,000 nodes on a switched LAN. 

As for congestion control, this is pretty much the same issue as UDP sockets or raw sockets: an application can send pretty much whatever it wants, regardless of the hop limit. The IETF-engineered applications tend to include some form of congestion control, e.g. a requirement of exponential back off when retransmitting, but then YMMV.

-- Christian Huitema



More information about the end2end-interest mailing list