[e2e] Random losses on very high speed networks
cannara at attglobal.net
Tue Jul 22 15:31:49 PDT 2003
Lloyd, I must only disagree on the basis of context -- "path" is here a
physical path, whether 1000 channels ride on it or not. Indeed errors could
occur just in one channel, due to hardware quirks, but that matters little to
the higher layers that then miss pkts on that channel. Now, if a mux/demux is
goofing up at the next layer, I agree, we mean logical path.
Lloyd Wood wrote:
> On Tue, 22 Jul 2003, Cannara wrote:
> > In fact, my use of "path" is here at the physical level, since
> > that's where Saad is going to be getting his fatal pkt errors.
> Which is why you should imo be saying "physical channel" instead.
> Which was the intended point of my previous post; to attempt to
> clarify that, you can still establish a point-to-point logical link
> over a multipath channel environment.
> > I agree with you that "bandwidth" is largely misused in the
> > networking and CS community -- anyone who takes a networking class
> > and still misuses bandwidth for capacity (at any layer) shoulda
> > flunked. :]
> hear ye, hear ye, O sigcomm bandwidth crowd.
> vaguely wondering offhand how much of Craig's 'Gigabit Networking' was
> devoted to Ethernet.
> > > On Mon, 21 Jul 2003, Cannara wrote:
> > >
> > > > Saad, if you mean other than LAN radio (wireless LAN...), then microwave and
> > > > optical links can be subject to losses that occur because of variations in the
> > > > path.
> > >
> > > a minor terminology nit, since microwave and optical links can often
> > > be multipath due to reflections - you mean variations in the physical
> > > channel here, rather than the (network) path.
> > >
> > > (Don't get me started on those comp sci types who wave
> > > physical 'bandwidth' around when they mean logical link capacity at
> > > some other layer.)
> > >
> > > L.
> <http://www.ee.surrey.ac.uk/Personal/L.Wood/><L.Wood at ee.surrey.ac.uk>
More information about the end2end-interest