—19—

functions to reference functions to special caching needs. In short, ISPs are engines for

innovation in markets we do not yet imagine.

49.

These ISPs thus could be important middle level competitors in the Internet

economy. They could provide an ongoing competitive threat to actors at their borders. In the

terms defined by Timothy Bresnahan, ISPs could become “vertical competitors” in an industry

marked by highly concentrated markets at each horizontal level.9Thus AOL, a traditional online

content and ISP, is now a potential threat to Microsoft in the operating system platform market.

This threat could not have been predicted three years ago. But by maintaining the fluidity of

borders, the net preserves the potential for new forms of competition.

50.

This layer of potential competition is especially important given how little we

know about how the broadband market will develop. The Internet market generally has been

characterized by massive shifts in the competitive center. Hardware companies (IBM) have been

displaced by operating system companies (Microsoft); operating system companies have been

threatened by browser corporations (Netscape) and by open-platform “meta”-operating systems

(Sun’s Java). As Bresnahan notes, we have no good way to know which layer in this chain of

services will become the most crucial. Thus, multiplying the layers of competition provides a

constant check on the dominance of any particular actor. Again, Bresnahan: “Far and away the

most important is that competition came … from another horizontal layer.”1 0Thus as he

IMAGE lem-les.doc02.gif

9See Timothy F. Bresnahan, New Modes of Competition: Implications for the Future Structure of the Computer
Industry
, June, 1998, http://www.pff.org/pff/microsoft/bresnahan.html.

10Id. at 6.

—20—

recommends, “one (modest) policy goal would be to make the threat of [vertical competition]

more effective.”1 1

51.

The architecture proposed by AT&T/MediaOne for their broadband cable service

threatens this vertical competition. By bundling ISP service with access, and by not permitting

users to select another ISP, the architecture removes ISP competition within the residential

broadband cable market. By removing this competition, the architecture removes an important

threat to any strategic behavior that AT&T might engage in once a merger is complete. The

architecture thus represents a significant change from the existing End-to-End design for a

crucial segment of the residential Internet market. Further, there is in principle no limit to what

AT&T could bundle into its control of the network. As ISPs expand beyond the functions they

have traditionally performed, AT&T may be in a position to foreclose all competition in an

increasing range of services provided over broadband lines.

52.

AT&T and MediaOne would achieve this change by bundling technologically.

The consequence of this bundling will be that there will be no effective competition among ISPs

serving residential broadband cable. The range of services available to broadband cable users

will be determined by one of two ISPs — @Home and RoadRunner, both of whom would be

allied with the same company. These ISPs will control the kind of use that customers might

make of their broadband access. They will determine whether, for example, full length streaming

video is permitted (it is presently not); they will determine whether customers might resell

broadband services (as they presently may not); it will determine whether broadband customers

11Id. at 18.

—21—

might become providers of web content (as they presently may not).1 2These ISPs will have the

power to discriminate in the choice of Internet services they allow, and customers who want

broadband access will have to accept their choice. Giving this power to discriminate to the owner

of the actual network wires is fundamentally inconsistent with End-to-End design.

53.

This technological bundling at the network level puts pressure on the principle of

End-to-End. These cable-owned-ISPs would thereby influence the development and use of cable

broadband technology. They would be exercising that influence not at the “ends” of the network,

but at the center. They are therefore shifting control over innovation, as Dr. Jerome Saltzer has

written, from a variety of users and programmers to a single network owner. This design defeats

the principle that the network remain neutral, and empower the users. It is the first step to a

return to the architecture of the old AT&T monopoly.

B.

The Costs of Violating Architectural Principles

54.

The costs of violating this fundamental principle of the Internet’s design are hard

to reckon. We simply do not know enough to know how sensitive the innovation of the Internet

is to changes in this competitive architecture. Obviously, in part the significance turns on how

the broadband market develops. But given trends as we can identify them now, the risks of

significant consequences from this design are great. We detail some of these risks below.

55.

The first is the cost of losing ISP competition. As we have argued, one should not

think of ISPs as providing a fixed and immutable set of services. Right now ISPs typically

provide customer support, as well as an IP address that channels the customer’s data.

Competition among ISPs focuses on access speed, as well as some competition for content.

IMAGE lem-les.doc02.gif

12These limitations are imposed by At Home Corporation. See @Home Acceptable Use Policy.
http://www.home.com/support/aup/(Visited November 8, 1999); At Home Corporation. @Home Frequently Asked

[made with GoClick]