[e2e] Bandwidth Estimation workshop

David P. Reed dpreed at reed.com
Tue Sep 30 06:21:56 PDT 2003

It's very sad to see that the vernacular use of "bandwidth" to mean 
"information rate" has been adopted by the networking 
community.  Information rate is a perfectly sound concept.

It's a mark of the ignorance of the community as a whole that it cannot 
think clearly and make clear distinctions.   I am sure that there are many 
professors of computer science and Ph.D.s in packet communications who may 
not even understand the technical distinction (since we don't teach signals 
and systems to CS people).

This would not be so bad, except there is no word or even an easy phrase 
for what "bandwidth" was coined to mean - which is the range of frequencies 
occupied by a modulated signal.   Those of us who work on RF and photonic 
systems (as well as those who deal with audio) are impoverished by this 
theft of a term, because for us, it is important to distinguish quite 
clearly between the information rate of a signal and its bandwidth in our work.

Bandwidth is measured in Hertz.   Information rate (or bitrate) is 
typically measured in bits/second (and I am considering a formal proposal 
to IEEE to call for a new unit, the Shannon, to deal with the lack of a 
named unit).   They are NOT at all equivalent.

It's as ignorant to say bandwidth when you mean bitrate as it is to say 
"light-year" when you mean a unit of time.   Except that physicists don't 
do that, and communications folks do.

Christian Huitema pointed out to me a week ago that we probably have to 
live with this loss of  a word.   So I suppose we'll have to coin a new 
word for the important old meaning.  I have proposed "frequespan" 
(FREE-kweh-span) to mean the range of frequencies occupied by a 
signal.   Of course, the voracious ignorance of CS professionals will 
probably colonize that one as well ...

More information about the end2end-interest mailing list