argues that a more general solution to the QoS problem is overprovisioning (7), and mo-
tivates that argument with strong evidence of the incentives of network providers to dif-
ferentiate between different applications. This, she argues, reduces "the incentives for a
potential innovator to develop new applications." Ibid., 7.

Van Schewick is the first researcher I know of to link QoS, overprovisioning, and end-
to-end, and I have drawn heavily upon her understanding in this context and throughout
this book.

Others are skeptical of the overprovisioning solution. For overprovisioning to work,
they argue, overprovisioning must exist at every point of the network. But that requires
coordinated action that is impossible in the context of the Internet. Thus, the solution is
an ideal without a mechanism to achieve it.

This skepticism is appropriate, but it does suggest another response. If the market
"naturally" would not choose overprovisioning, but overprovisioning would nonetheless
be a productive response, then we may well have identified a role for government. I de-
scribe this a bit more in Chapter 12.

[3-70] For examples of QoS technologies, and a discussion of QoS generally, see Joe
Lardieri, "Quality of Service: Which Flavor Is Right for You?," _Telecommunications_
(August 1999): 53; Dave Kosiur, "Directing Traffic with a Touch of Class," _PC_Week_
(March 23, 1998): 91; David B. Miller, "Quality of Service: Directing Data Through the
Network," _ENT_ (October 6, 1999): 22.

[3-71] See Gilder, _Telecosm,_ 158-164.

[3-72] The critical change necessary for this "fibered" network to achieve this level of
capacity is the emergence of optical routing technologies. The emerging bottleneck in
the network is the relatively slow speed of electronic switches. For an introduction, see
"The Ultimate Optical Network" (Special Report), _Scientific_American_ 284 (January
2001): 80.

[3-73] See, e.g., Bill Frezza, "Telecosmic Punditry: The World Through Gilder-Colored
Glasses," _Internet_Week_ (December 4, 2000): 47; Rob Walker, "The Gildercosm," _Slate_
magazine (September 11, 2000), available at http://slate.msn.com/code/MoneyBox/
MoneyBox.asp?Show=9/11/2000&idMessage=6030 (Gilder runs "against the current
wisdom that sees bandwidth shortage as a problem"); Julian Dibbell, "From Here to In-
finity," _Village_Voice,_ September 5, 2000, 65 ("It takes either profound sloth or transcen-
dent faith to persist in voicing such breathless sentiments."). For a more favorable review,
see, e.g., Blair Levin, "Review, _TELECOSM:_How_Infinite_Bandwidth_Will_Revolution-_
_ize_Our_World,"_Washington_Monthly_ (September 1, 2000): 54.

[3-74] E-mail from Timothy Wu to Lawrence Lessig, February 16, 2001.

[3-75] This claim depends upon the assumption that the value of the network activity was
higher than the telephone activity it displaced. That assumption about social value
therefore has little relation to the actual costs.

Chapter 4

[4-1] Technically, some of these functions could be provided if the program simply pro-
vided perfectly transparent representations of the "application program interfaces," or
APIs. In some contexts this distinction will be important, but I ignore it here. A suffi-
ciently skilled programmer could also learn something about compiled code through re-
verse engineering. The success of this would depend upon how well structured the
original code was.

[4-2] Ceruzzi, _A_History_of_Modern_Computing,_ 108.

[4-3] For a brief history of Unix, see William Shattuck, "The Meaning of UNIX," in _The_

[[278]]

p277 _ -chap- _ toc-1 _ p278w _ toc-2 _ +chap+ _ p279

v?