Internet and Internet Telephony

We would be remiss not to discuss the Internet in a book devoted to competition in telecommunications. The Internet offers a wide variety of new services, as well as new ways to provide old services like telephony. Internet telephony, as its name indicates, refers to telephony over the Internet; unlike a traditional phone call, for which a circuit is opened and dedicated to a single conversation for the whole length of the call, messages sent over the Internet are decomposed into tiny data packets, which may or may not take the same route, and which are reassembled at termination. Internet telephony uses the transmission control protocol/Internet protocol (TCP/IP), which more generally supports transmission of packets over the Internet regardless of their application (voice, audio, video, . . . ).

Internet telephony currently has a low market share, and arguably this market share is inflated by favorable regulatory treatment. The low market share in part results from the low quality of calls; Internet telephony, like other premium services such as videoconferencing and unlike e-mail, requires very low delays of transmission over the Internet. Such low delays are not yet guaranteed, and one of the key challenges for the development of Internet telephony is precisely the definition of protocols and interconnection agreements involving prioritization of premium services that will enable networks to promise their customers an adequate quality of service. Most experts predict a rapid growth of Internet telephony in the next few years. This section briefly discusses the future of the Internet more generally, as well as two key challenges currently facing its development: broadband access to the home and interconnection. These challenges raise issues of one- and two-way access, respectively. Their full analysis lies outside the scope of this book, and we will content ourselves with a description of the main issues.

Broadband Access

Currently, most consumers can connect to the Internet from their home or small business premises through dial-up access at very low speeds. Broadband access to the home would provide speeds of transmission, say, one hundred times faster than current speeds, as well as an "always-on" capability.

Several technologies are envisioned to break this "bandwidth-to-the-home bottleneck":
  • DSL technologies: By installing new-generation modems at the customer's premises and at the location of the first switch, the owner of the copper line or else an entrant having "unbundled access" to the copper line can substantially improve its performance. ("Unbundled" refers to the fact that the entrant purchases only access to the naked copper line and not other services such as switching.) Several operators are currently performing commercial experiments with these technologies. We will come back shortly to the issue of local loop unbundling (LLU).
  • Cable: Cable transmission facilities, which historically have been used to transmit content to the home, usually have only one-way capability (cannot offer interactivity). But once cable modems are set in place, they are able to provide two-way, higher bandwidth capability.
  • Fiber to customer premises: Bringing fiber to the home would be an ideal system of delivery to a fixed location, with data delivery rates much larger than those for DSL technologies. Alas, the cost of putting fiber to the home is very high, and this solution is unlikely to be adopted in the short run. (Of course, businesses with large usage already have fiber to their premises.)
  • Power line: There is currently some experimentation to modify power distribution networks in order to provide customers with high-speed access to the Internet through electricity wires. Like fiber to the home, this is not a realistic possibility in the short run.
  • Radio spectrum and satellites: A number of wireless solutions are currently being considered. For example, some consortia are considering using a satellite constellation to provide "local access."
Let us return to the issue of local loop unbundling. Simplifying a lot, there are three forms of access to bandwidth that can be offered to an entrant:
  1. Rental of naked copper line: The entrant rents the copper line from the home to the first switch from the incumbent operator and collocates with the operator so as to be able to install its modems. The bandwidth then belongs to the entrant, who can make the commercial use of it that he wishes (offer his own services or rent bandwidth to providers of final services). Presumably, the incumbent and the entrant then compete in packages or bundles, in the same sense in which two cable services providers (e.g., wireline and wireless) compete in bundles for the customer. The customer will choose her supplier, who then will install the modems and will provide a range of services using the bandwidth (produced in house or outsourced).
  2. Exclusive bitstream access: The incumbent installs the modems on behalf of the entrants (as well as for himself). The entrants then do not rent only access to the copper pair, but rent the entire bandwidth. There is no competition in building facilities, but there is competition in bundles of services. Thus the key difference with the first option is that the investment in extra facilities (in particular, the modems) is here always borne by the incumbent, whereas it is borne by the entrant (provided the consumer chooses the entrant) in the first option.
  3. Nonexclusive bitstream access: As under exclusive bitstream access, the incumbent keeps a monopoly on the building of new facilities, but instead of renting the entire bandwidth to a single entrant, he sells pieces of this bandwidth to different entrants at some access price per unit of bandwidth. This solution allows an entrant who does not wish to offer a full range of services to contract directly with the customer.
The choice of regulatory framework will be crucial for the development of broadband access to the home. Key issues include the choice of technology (xDSL technologies are still improving) and their compatibility with the services that entrants and incumbents desire to offer; the optimal sharing of the investment risk between incumbent, entrant, and customer (in view of rapid technological progress in these and alternative technologies and of an important uncertainty about demand for the new services, economic depreciation ought to be large); the design of regulatory commitments against takings of the new facilities; the definition of proper access charges (cost oriented versus demand-and-cost oriented, measurement of cost, nondiscrimination rules, relationship between local loop rental and consumer's monthly subscriber charge, etc.); and relationship to universal service. 

Interconnection in the Commercial Internet

Probably the largest stakes in telecommunications today lie in the Internet. Yet, little is known about the future industrial organization of the Internet. Indeed, the biggest part of the initial network, the NSFNET, was privatized only in 1995, and thus the commercial era is just beginning. Until recently the Internet community was largely one of engineers working cooperatively to take the Internet off the ground. Nowadays, financial stakes are huge, and the Internet is turning into a fascinating commercial battleground.

Again, we content ourselves with a brief description of the issues, starting with a description of the players. End users include residential users and businesses, who have access to the Internet either through dial-up (over the phone line using modems) or through dedicated access. On the other side lie web sites, which provide a wide variety of free or fee-based content as well as offerings of services (e-commerce, . . . ). In between can be found a host of intermediaries. Some intermediaries provide users with guidance as to whether to connect, what to buy, and so forth: search engines, portals, infomediaries, . . . Other intermediaries provide transmission services: internet service providers (ISPs), internet backbone providers (IBPs). We will focus on the latter, keeping in mind that the dividing line between the various players is not always clear-cut: For example, America Online (AOL), an ISP offering Internet access to residential users, also offers content as well as search capabilities (for example, through its acquisition of Netscape, which produces a browser with search-engine capability).

Internet backbone providers direct traffic over large regions of the world using long-haul fiber-optic cables. IBPs connect to each other at multiple points under the so-called "peering agreements" (see following paragraphs). IBPs pick up the traffic generated by ISPs (as well as that of their own customers) and carry it over long distances. They also have the most sophisticated routing of all Internet players.

The Internet is a network of interconnected networks. Indeed, one of the main appeals of the Internet is its current almost ubiquitous connectivity: From almost any point (URL address) in the network can be sent messages to almost any other point. One may wonder how a network of 7,000 ISPs and 4 to 50 IBPs (depending on the exact definition of IBPs) can offer such ubiquitous connectivity. The basic structure is hierarchical.

IBPs "peer" with each other. In so doing, they accept for routing all traffic that is destined to their own customers, the customers of their customers, and so on. Peering used to occur at public peering points, NAPs (network access points), or MAEs (metropolitan access exchanges). The slow expansion of the capacity at these points (Internet traffic grows at a rate of up to 1000% a year) has led IBPs to turn to private peering, that is, to exchanging traffic pairwise at a number of bilateral interfaces. The importance of public peering points, where an arbitrary number of networks exchange traffic, is waning. IBPs impose a number of conditions to accept each other as peers: number of points of interface, national high-speed network, and so on. Currently, peering arrangements are of the bill-and-keep type; that is, each peer terminates without charge the traffic originating with other peers. This feature is probably a leftover of the transition process. One may wonder whether IBPs will keep running their two-way interconnection arrangements through bill-and-keep.

IBPs do not make money from their peering relationships. To cover the huge investments they have made in infrastructure, they charge their customers, who in turn charge their own customers. Charges often are related to the capacity of the link between the network and its customer, but can also depend on usage. Thus the Internet can be seen as a pyramid, in which monies are collected at the bottom.

To be certain, the organization of the Internet is not purely hierarchical. For example, it may make sense for two ISPs in the same city, such as ISPs A and B, to exchange traffic directly (engage in "secondary peering") rather than let their mutual traffic move up and then down the hierarchy. Such sideways interconnections do not upset the hierarchical nature of the Internet.

The design of interconnection arrangements is crucial for the future of the Internet. In the short term, it conditions the prices charged to dial-up and dedicated access users and to web sites, and thereby the use and organization of the network. In the "long run" (a few months in the Internet world, owing to the growth of traffic and technological progress), it determines the networks' incentives to build up their capacity and to cooperate.

The framework for two-way interconnection for traditional voice telephony provides a number of insights that will be useful for our understanding of Internet interconnection. However, as it stands, it is inappropriate, in that it does not reflect the specificities of the Internet. These specificities not only are technical (packet switching versus dedicated circuit), but they also have important economic dimensions. For example, unlike voice telephony, the party who requests the message may receive rather than send this message, as is the case when a user downloads a web page. Such specificities require a careful consideration of the interaction between the wholesale market (interconnection arrangements) and the retail markets (pricing to end users and web sites and commercial interaction between these).

A second important issue related to interconnection is the development of premium services. Premium services on the Internet (IP telephony, video on demand, videoconferencing, etc.) require low delays in packet transmission and therefore a higher quality of service throughout the network than is currently observed. Several scenarios may be envisioned. First, some large Internet operators may develop proprietary standards and offer such services on a limited basis (between their customers), hoping that the lack of ubiquity will be mitigated by tipping, or at least that the proprietary offering will create a comparative advantage. Second, networks may agree on standards and two-way access charges for the premium services to attempt to achieve ubiquity. The incentives for cooperation and the design of two-way access charges for premium services are important topics for research.

Alternative Policies | Multiple Bottlenecks and Two-Way Access

Two-way-access policies have historically received little attention in the policy debate, except perhaps for international telecommunications. They are bound, as we have seen, to become more and more prominent over time. The policy debate and the theoretical analysis have by and large focused on particular approaches, namely, those of regulated or privately negotiated determination of reciprocal access charges. While these are reasonable paradigms, others are worth investigating as well. Without any pretense at exhaustive and careful analysis, we record two alternative possibilities.

Making the Receiver Accountable

A striking feature of interconnection frameworks in which the calling network pays the termination charge is that if termination charges are not set reciprocal, call receivers do not internalize an externalitythe termination charge paid by other networksgenerated by one of their decisionstheir choice of network.

Let us entertain the possibility that networks freely set their own termination charges and have call receivers pay the termination charge set by their elected networks (this termination charge being the same for all calls, whether they originate on or off net). Networks would then be unable to "tax" other networks through a high termination charge.

A possible objectionand a standard one for motivating the absence of retail charge on the receiving sideis that call receivers do not like to pay for being called; this is particularly the case for nuisance calls. But there are ways of accommodating this concern. Suppose for example that the call receiver is charged the difference between the termination charge set by his network and a benchmark termination charge (the calling network would then pay the benchmark termination charge). This benchmark termination charge could be derived from a cost model, or else (and perhaps more in line with the idea of using market-determined access charges) be an average of termination charges set by other comparable networks. In this case, the call receiver would pay nothing for receiving calls if this network adopts the benchmark termination charge. Yet, the call receiver would still be fully sensitized to changes in his network's termination charge.

What would be gained and lost relative to the institution of a jointly set reciprocal access charge? The potential benefit is that a network can more easily reflect its termination cost specificity and transmit the corresponding signal to customers. So, for instance, a mobile network with sparse (excess) capacity can translate its marginal cost of termination into a high (low) termination charge. Similarly, a wire-based network would not set the same termination charge as a wireless network. But does this policy bring any benefit? Let us entertain two possible hypotheses concerning the elasticity of demand for terminationthat is, the sensitivity of receiving minutes to a per-minute termination charge imputed to the subscriber.

Suppose first that the subscriber has no control over the volume of calls he receives: His demand for termination is inelastic. The termination charge then has no impact on call termination, and a unit increase in the termination charge is equivalent to an increase in the fixed fee (monthly subscriber charge) equal to the volume of calls received. Note, though, that this impact on the "generalized monthly subscriber charge" is not uniform across consumers. While the payment of a termination charge by the customer has no effect on the volume of calls received, it allows the phone company to have a better measurement of phone usage and thus of the customer's willingness to pay for receiving calls. This knowledge may enable the phone company to engage more effectively in price discrimination (although this discrimination may be complex when the fraction of calls that are nuisance callsand thus do not raise the willingness to pay for being connectedvaries across customers).  For example, customers who mainly receive calls would tend to migrate away from high-termination-cost networks, and this migration would create a socially beneficial reallocation of consumers.

Second, let us assume that the demand for termination is somewhat elastic. In practice, consumers may give their phone number to a restricted set of acquaintances and require that their number be unlisted, and they may abbreviate the calls they receive. Making customers accountable, then, may have costs and benefits. The benefit is that consumers are induced to receive fewer calls and to abbreviate calls when their network's termination cost is high. A potential cost of making customers accountable is that networks would not fully internalize the externality on callers. A network's high termination charge induces its subscribers to abbreviate calls. The externality on callers belonging to the same network is internalized by the network, but not that on callers subscribing to the rival network. We should be careful not to draw conclusions, however, since we have not conducted an analysis of industry behavior in this framework analogous to that developed for reciprocal access charges.

Multiple Terminating Lines

The possibility that not only local networks but the very connection to the customer's home may be duplicated raises new possibilities as well as questions about the likely industry behavior and performance. This point has been known for a while with regard to call origination: For example, a large user can be connected by, and actively subscribe to, two operators, and use real-time least-cost routing to decide which network to use at each instant. But it is evident that similar possibilities arise at the termination level. Suppose that a customer's home is connected both by a telephone company and by a cable company that has upgraded its line to carry telephony. Whether the customer subscribes to one or both networks for originating calls, there is now a potential choice of operators for terminating calls. Networks originating calls, or the receivers themselves if they are charged for call termination, can select the least-cost termination. Again, it would be worth investigating the implications of such competition for termination for industry behavior and performance.