[e2e] Bandwidth Estimation workshop

[Michael B Greenwald]

   Tue, 30 Sep 2003 15:20:34 -0400
   "David P. Reed" <dpreed at reed.com>

I agree with you that it would be better to have two distinct terms.
However, I disagree that having a single term necessarily implies that
people are unable to understand the distinction between "bandwidth"
and "bandwidth.  And I disagree that simply because people use a term
in a new sense you can't still use it for the old sense --- as long as
the distinction is clear in context.  (It would still be better to
have distinct terms, but civilization won't end if we stick with one.)

No one thinks the two bandwidths refer to identical concepts (the fact
that you have to multiply one by something with units like
bits-per-cycle is a dead giveaway that they refer to totally different
things).

Note that we are not the only community to overload terms in a way
that leads to confusion.  Physicists did it, too.  Consider "pound".
It is a unit of mass ("avoirdupois (spelling?)  pound") and a unit of
weight ("pound force"), and the two are clearly not identical
(although at the earth's surface they have equal magnitudes: a one
pound mass has a weight of one pound).  And also note that having
different terms for mass and weight doesn't avoid some people
confusing the two (if you have elementary school kids, and you've
tried explaining to their teachers that weight is a force and not a
mass then you know what I mean).

You may think I'm cheating because the *full* names for the two
"pounds" are distinct (although I never use the full names, and don't
remember how to spell "avoirdupois").  Then consider "degree", or
"velocity" (We all know the normal definition.  We also know that it
is used vernacularly as a scalar quantity. However, it is also used as
a scalar in a technical sense; it is also "the speed at which the
detonating wave passes through a column of explosives, expressed in
meters per second") or "viscous force" (apparently there are two
meanings, one a measure of force per volume, the other a measure of
force per mass).

It would be great if all of these concepts had distinct terms, but
they don't, and yet people still make appropriate distinctions.  

So I agree that there should be two distinct terms for bandwidth and
bandwidth, but I don't agree that you can't talk about old-fashioned
bandwidth anymore.  And for people who confuse the two concepts, I
won't guarantee that having distinct terms for bandwidth and bitrate
will help any more than "weight" and "mass" help in *their* domain.

Also:

   At 11:27 AM 9/30/2003, Constantine Dovrolis wrote:
   >On a personal note: I teach CS-networking and my students know very well
   >the distinction between physical-layer bandwidth and network-layer
   >bandwidth.
   
   This is a distinction which I have not heard.   Do you have a textbook 
   citation, or perhaps a citation of a survey paper, that provides such a 
   distinction?

   I ask because "information rate" is well defined at both layers in 
   bits/second, yet you would claim that the distinction depends on the layer.
   
I think you are misinterpreting what Constantine Dovrilos wrote.  I
don't believe that he meant that the distinction was dependent, in any
way, on the "layer".  He was simply using "physical layer" as a
shorthand for EE, physics, (i.e. domains in which "bandwidth" is used
in the old, conventional sense) and "network layer" for CS,telecom
(i.e. communities who have adopted (co-opted?)  "bandwidth" to also
mean information rate.)

I believe (but may be wrong, and don't mean to speak for him) that he
*is* arguing with you.  But his argument is simply that people are
able to distinguish between "bandwidth" (old) and "bandwidth" (new)
and therefore it is not *urgent* to crusade for new terminology
immediately.  His argument has nothing to do with spurious
distinctions based on layers.  He is simply claiming that his students
would be able to carry on a conversation with you without getting
confused as to which bandwidth you were refering to at any given time.
(I'll assume he is a reasonable judge of his students' abilities.)

   And here in my research group, we construct transport networks that operate 
   in the RF domain, doing routing and switching using the medium and spatial 
   modulation and filtering.   That "network layer" that we construct has a 
   bandwidth that is measured in Hertz, which has nothing to do with the 
   information rate, which is a function of the spatial sampling density and 
   energy density (according to a law that looks a bit like a Shannon capacity 
   theorem).
   
   My point is that by deliberately conflating the two meanings of bandwidth 
   and defining them in terms of an arbitrary layering convention (network vs. 
   physical), this pedagogy ennobles a pernicious form of ignorance - one in 
   which systems such as the ones we are building here cannot even be 
   described, because the terms taught prevent students from being able to 
   express them.