TABLE OF CONTENTS INTRODUCTION I. THE AGREEMENTS FORMING THE INTERNET A. Four Ways to Communicate: Peering, Transit, Paid Peering, and Intranetwork 1. Peering: What's a Few Bits Between Friends? 2. Transit: The Cost of Doing Business 3. Paid Peering: When it Absolutely, Positively Has to be There in Twenty Milliseconds 4. Intranetwork Traffic: Not Exciting, but Free B. Technologies and Trends in Data: More Revenue and Less Expense for ISPs II. ISPs INSTITUTE DATA CAPS A. Cataloguing Data Caps: Size, Penalties, and Staying Power. B. Why Cap? Costs and Congestion Money and Power III. REGULATORY SOLUTIONS: ANTITRUST, NET NEUTRALITY, AND PRIVATE AGREEMENTS A. Antitrust Regulation 1. Possession of Monopoly Power: Too Much Success 2. Willful Maintenance of Monopoly Power B. Net Neutrality: Vague Rules with an Uncertain Effect 1. Do Data Caps Block Access? Only if the FCC Wants Them To 2. Net Neutrality as a Defense: The FCC Made Me Do It! C. Individual Agreements: Ending Data Caps One at a Time IV. CONCLUSION: MORE THAN JUST TV--DATA CAPS AND THE INTERNET'S EXPONENTIAL GROWTH INTRODUCTION
The advent of near-ubiquitous high speed Internet access has prompted substantial regulatory, legislative, and academic scrutiny over how much leeway Internet service providers ("ISPs") should have to offer preferred treatment for services like Hulu. Some have compared this kind of preferential treatment to creating an Internet "fast lane." (1) More attention should be devoted, however, to the limits ISPs are now placing on their customers' aggregate monthly Internet usage, also known as data caps. If other types of regulations can be characterized as governing speed limits on the information super highway, data caps will ultimately determine customer mileage.
This paper examines current data cap regimes, probable effects on network usage, and what, if any, action regulators can and should take. Part I analyzes the interconnection agreements that, in large part, determine the incremental cost ISPs pay for the data that customers download. This section finds that the most data-intensive network uses are also frequently the least expensive for ISPs and often actually serve as profit centers. Part II looks at the data cap policies imposed by various ISPs and considers the motivation behind them. Given the low incremental cost of data, caps hardly seem to be a price control mechanism. Some evidence suggests that data caps may be a price-gouging tool similar to overages on cellphones. Moreover, data cap policies seem to have the intention of dissuading customers from moving their TV viewing from traditional multichannel video programming distributors ("MVPDs"), which include cable providers, to online video services. Part III considers potential mechanisms for government regulation of data cap policies. Specifically, it proposes that the Department of Justice bring an antitrust action against cable ISPs. Additionally, the recent net neutrality rulemaking by the Federal Communications Commission ("FCC"), if upheld by the courts, also creates new opportunities and challenges for regulating all ISPs, not just cable, and individual settlements may provide quick regulation on a case-by-case basis. Part IV concludes that there is a need for regulatory pressure. If the capacity of consumer Internet access does not grow with the speed of that access, the exponential growth in data usage that has driven the information economy may falter.
THE AGREEMENTS FORMING THE INTERNET
ISPs are in the business of sending information to and from customers. To do this, the ISPs--owners of the last mile of cable to a customer's house--contract with backend service providers, owners of high-bandwidth interstate and international connections, to transmit data across the globe. These contracts take a variety of forms, and the answer to even the most fundamental questions--such as "which party pays?"--will change depending on the circumstances. Understanding the agreements and technologies governing this flow of data is necessary to understanding the motivations behind data cap policies. However, unlike the morass of statutes, regulations, and cases long governing telephone interconnection and transmission, (2) the agreements between an ISP, other nearby networks, and the backend networks that connect to networks nationwide are almost entirely unregulated. (3)
Internet communication is fundamentally different than traditional phone service. (4) When a phone call is placed, a company's circuit is monopolized and no other customer can use that resource. (5) When data is sent over the Internet, however, it is shaped as a packet and shares the connection with other users. (6) This technological novelty led to a diverse array of proposed and practiced billing structures. (7) The structure negotiated by the ISP with the backend providers determines how much sending and receiving data will cost--or in some cases, profit--the ISP. (8)
Four Ways to Communicate: Peering, Transit, Paid Peering, and Intranetwork
ISPs and backend providers have begun to settle on the unit of accounting that will be used to calculate traffic bills. While this seems like a simple first step, experts previously suggested diverse accounting methods. (9) International telephony settlement practices provided one model, but such agreements were inapt because the Internet packet structure does not have an analogous "call-minute" to base billings on. (10) Proposed units of billing under this model included transmission control protocol/internet protocol ("TCP/IP") call-minutes and billing based on the number of packets sent, with variants including measures of the size of the packets sent. (11) However, as of 1999, no single model was in widespread use. (12)
Rather than adopt a complex accounting unit that required detailed examination of user behavior, network providers (both ISPs and backend providers) have trended towards billing for the capacity of a connection, measured in bits per second. Although the secrecy of interconnection agreements makes it difficult to determine exactly how widespread is adoption of this or any particular billing unit, (13) one primer on the topic suggests that the common practice is for companies to reserve internetwork speed in bits per second (capacity), rather than to bill for total usage. (14) This assumption, that capacity-based billing is the standard approach to interconnecti on agreements, is consistent with the public information in the recent Comcast-Level 3 dispute. Comcast, an ISP, demanded that Level 3, a backend service and content delivery network provider (described in more detail infra), pay for its connection based on how many interconnection ports (which offer a fixed bandwidth in gigabits per second) were used. (15) Capacity-based billing is also consistent with Canadian regulations that govern interconnections between small and large ISPs. (16)
In order to send information outside of its network, i.e., to increase capacity, an ISP may enter into three types of agreements: Peering, Transit, or Paid Peering. An ISP may also avoid the need for extra network communication by using intranetwork resources. (17)
Peering: What's a Few Bits Between Friends?
In some cases, an ISP will avoid paying for data by entering into "peering" agreements. Under these arrangements, both networks (ISP and backend) will interconnect and agree to forego any regular account settlement payments on the assumption that all sides are receiving roughly equal value from the arrangement. (18) Any data sent from one network, which is to be delivered to (or "terminated" on) a peering partner's network, is transmitted free of charge. However, no data will be delivered to any networks which are not party to the peering agreement, even if one of the peering partners purchases data from that network provider. (19)
The main advantage of this kind of arrangement is cost. In a peering agreement, the only costs a network will incur are for the equipment and physical transmission capacity. (20) There is no transactional cost related to billing or measuring peak bandwidth demand, and there is no marginal cost for sending lots of data. (21) As such, peering agreements are popular between major data carriers, or "tier-1" networks, such as Sprint, AT&T, Verizon, and Level 3, who are able to reach every other network without paying a settlement. (22) These carriers are also incentivized to make connections with as much bandwidth as technically possible, thereby diminishing the chance that packets will be dropped. (23) Small and regional ISPs may also enter into peering agreements with other small networks, where roughly equivalent amounts of data are exchanged between the two networks, (24) or with content delivery networks (CDNs), which locate servers for media rich applications like Hulu, (25) iTunes, (26) and Netflix (27) near ISP servers, reducing the time to deliver this content to end users. (28)
Peering arrangements are limited by their inability to reach every network on the Internet. It is a nearly impossible for one network to connect with all the hundreds of thousands of other networks on the Internet. Even large networks such as AT&T must, in a way, rely on small networks that purchase access from the large network. (29) Significant changes in data usage by either side in a peering arrangement may also lead to "depeering," resulting in dramatic cost shifts for both sides, occasional calls for regulatory action, and network interruptions. (30)
Transit: The Cost of Doing Business
Transit is the opposite of peering. Rather than transmit data freely between agreeing networks, one network--for instance, a regional ISP wishing to gain access to a nationwide network--will purchase bandwidth from the second network at a recurring fee. (31) A transit customer will limit its interconnection speed as much as possible instead of attempting to connect at the fastest...