Interconnection of different networks is critical to ensure a competitive communications market. It is fundamental for service providers to ensure that their users have the ability to connect with users of any other network or service provider. As the Internet expands and becomes more geographically ubiquitous and as traffic increases, more efficient IP-based Internet interconnection will be required. This is especially relevant for developing countries, where lack of interconnection facilities means that Internet traffic originating there is mostly subject to “tromboning” (that is, using international transit facilities to deliver local traffic).13 See infoDev and ITU, “ICT Regulation Toolkit, Module 2: Competition and Price Regulation,” sec. 4.8.1, The Role of Internet Exchange Points, http://www.ictregulationtoolkit.org/en/Section.3571.html. Policies to facilitate national or regional Internet exchange points (IXPs), the physical infrastructure where Internet service providers (ISPs) exchange Internet traffic between their networks, will play a crucial role in ensuring more efficient and cost-effective Internet interconnection in these countries.

Similarly, as the transition toward IP-based NGNs proceeds, questions will arise regarding the manner and terms under which IP-based interconnection will take place between different types of networks and at different functional levels of the network. Especially relevant are issues relating to future wholesale charging mechanisms that may apply to converged broadband networks. The following sections address current trends and expected regulatory developments relating to these issues.

3.4.1 Internet Interconnection and IXPs in Developing Countries

Historically, the exchange of Internet traffic has been focused in developed countries. In the early years of the Internet, traffic was routed and exchanged mainly in the United States. As Internet access has expanded and the amount of content available has increased, the exchange of Internet traffic has been distributed to other developed countries in Europe and Asia through the creation of national and regional IXPs (Kende 2011, 25). In the case of developing countries, while IXPs have been progressively implemented and peering is occurring at the national level, the amount of traffic that is exchanged within developing regions is still very small. Most of the traffic is still hauled out of the region for switching and then sent back into the region for delivery.

IXPs in developing countries are important for Internet interconnection for several reasons. By providing an interface for the exchange of local and regional traffic, IXPs facilitate a more efficient and cost-effective management of international bandwidth. Because of their small volume of traffic, ISPs in developing countries mostly have to rely on transit agreements, since the largest providers do not have incentives to enter into shared-cost peering agreements with them. Due to the charging mechanisms for international Internet transit, this means that the developing-country ISP will ultimately bear the costs of outbound and inbound traffic. Local peering through IXPs at the national or regional level helps to resolve this problem and reduces the costs of Internet access for consumers in developing countries.

More local interconnection, in turn, allows for the provision of more reliable services, with lower latency that then can support multiple, innovative, time-sensitive applications. For example, for African ISPs, tromboning adds an estimated 200 to 900 milliseconds to each transmission. This added latency can impede the development of new services, such as Internet telephony, streaming audio and video, video conferencing, and telemedicine. By interconnecting at a local IXP, two ISPs (located near to each other) can overcome this problem and route traffic to each other’s networks in 5 to 20 milliseconds.14 infoDev and ITU, “ICT Regulation Toolkit, Module 2: Competition and Price Regulation,” sec. 4.8.3, Internet Exchange Points in Africa, http://www.ictregulationtoolkit.org/en/Section.3573.html.

As noted, IXPs are now being implemented in some developing countries. Before 2002, there were only two IXPs in Africa, with this number increasing to 10 by 2003. By December 2010, there were 20 IXPs distributed among African countries. While this represents significant progress, the great majority of Internet traffic from Africa, around 85 percent, still relies on connections to Europe; just 1 percent of the traffic being exchanged stays within the region.

From a regulatory perspective, a series of barriers can hinder Internet interconnection and the establishment of IXPs in developing countries (McLaughlin 2002). As discussed later in this chapter, Internet interconnection has developed under market-based mechanisms and without the need for regulatory intervention. However, regulators’ attempts to extend their mandates to encompass Internet interconnection may result in unwarranted regulation and create disincentives for the deployment of IXPs. These include, for example, legal restrictions that prohibit the deployment of nonregulated ICT facilities, such as IXPs. Unduly restrictive or burdensome licensing regimes may also limit the deployment of IXPs. Similarly, exclusive rights for the provision of international connectivity, which some countries maintain, can also impede efficient Internet interconnection.

In some cases, lack of appropriate regulation of the inputs required to implement effective IXPs, such as national backbone connectivity, may result in above-cost rates for wholesale services. For example, high costs of leased lines can significantly affect an IXP’s viability. In addition, deficiencies in regional broadband connectivity play a role in the continued low levels of intraregional Internet traffic exchanged in developing regions, like Africa (Stucke 2006). Lack of relevant local content also affects the extent to which traffic is peered within national or regional IXPs.

Box 3.2 presents a case study of the implementation of the first IXP in Kenya, which illustrates some of the legal and regulatory difficulties outlined above and how they were overcome.

3.4.2 IP-Based Interconnection: Wholesale Charging Arrangements

Despite the increasing physical and logical integration between legacy public switched telephone networks (PSTNs), public land mobile networks (PLMNs), and all-IP networks, two separate models are still typically used for exchanging traffic in these networks. Internet traffic is exchanged using IP-based interconnection and relies on privately negotiated peering and transit agreements. PSTN and PLMN traffic, however, may be exchanged using a combination of switched and IP-based interconnection, but it is normally subject to regulation and typically falls within two main wholesale charging arrangements: calling party network pays (CPNP) and bill and keep (BAK).

As convergence toward NGNs advances, these differences create potential arbitrage opportunities between regulated and unregulated services and lead to potential competitive distortions (BEREC 2010, 8). Regulatory authorities are therefore considering what reforms in wholesale charging mechanisms, if any, should be implemented at the national level for termination services to enable IP-based services and broadband further. While it is not clear which wholesale charging arrangements will prevail, some authorities are expecting that a uniform wholesale charging mechanism for IP-based interconnection may emerge in the future.

3.4.3 Current Wholesale Charging Arrangements

The majority of countries around the world use CPNP for PSTN-PLMN interconnection at the wholesale level. Under this system, the originating network is required to pay a charge, generally per minute or per second, to the terminating network for the traffic exchanged. An alternative approach is BAK, which is used for PLMN in countries such as the United States, Singapore, and Canada and is a system where interconnecting operators generally do not charge each other for terminating calls. These terms are equivalent to negotiating termination rates equal to zero and typically include reciprocity obligations, meaning that the same terms are applicable to both parties to the agreement. Under BAK, the costs associated with call termination may in some cases be recovered from the service provider’s own subscribers as that provider sees fit—for instance, by levying a charge for calls received.

Efficient IP-based interconnection in the Internet has been achieved for the most part without the need for regulatory intervention. Since no single entity has the ability to connect to all of the networks that form the worldwide Internet, a series of indirect interconnection (transit) and direct interconnection (peering) arrangements have developed to ensure that traffic will reach its intended destination. In Internet interconnection, the combined framework of transit and peering, together with the IP packet routing protocols, remove the a priori case for regulation based on the termination monopoly present in PSTN-PLMN interconnection under CPNP systems. For example, if an ISP denies direct interconnection (peering) to another ISP, the latter ISP is generally capable of accessing customers of the former, although at different costs, as long as it has an indirect (transit) agreement with a third party.15 The point that peering and transit arrangements are demand-side substitutes has recently been made by the European Commission in a case involving the Polish regulatory authority’s proposal to regulate these services as separate relevant markets. See European Commission (2010a, para. 36). This same result is not generally possible in the circuit switched environment. If the PSTN-PLMN provider refuses interconnection, competitors generally cannot terminate calls to its subscribers.

3.4.4 Future Charging Mechanisms

In the long run, the differences in interconnection charging arrangements will not likely be sustainable or efficient in a converged NGN environment, where more traffic will be IP based. Price differences between regulated and unregulated interconnection services result in arbitrage opportunities and potential market distortions. Therefore, a uniform wholesale charging system may be needed for future NGN interconnection. This could be based on the Internet economic model (Marcus and Elixmann 2008, 114), the PSTN-PLMN model, or some third option resulting from a combination of both (European Commission 2009a, 32). Others emphasize that, although NGNs and the Internet use IP as a common technology and are converging in the marketplace by offering similar or substitute services, they are organized differently and so remain separate and distinct, even though they share the same transmission infrastructure, such as fiber networks (Tera Consultants and Lovells 2010, 79–92). Consequently, it is argued that the two types of networks will not converge since the Internet is a collection of “open networks” and NGNs are a collection of “closed” networks (that is, packets cannot be allowed across the interconnection point unless they are authorized), and hence there is no convergence-based argument in favor of a uniform charging system for NGNs based on BAK.

Despite this, there are some early indications that future wholesale price mechanisms may resemble IP network pricing, that is, PSTN-PLMN per minute or per second pricing may migrate to pricing based on barter arrangements (for example, BAK) or on capacity-based interconnection (CBI). A recent attempt by the Polish regulatory authority to lead regulation in the other direction (that is, regulating the terms, conditions, and prices for Internet peering and transit services using tools similar to those applied to PSTN-PLMN) met with significant opposition from the European Commission (EC) and was eventually discarded in March 2010 (European Commission 2010a).

Similarly, the Body of European Regulators of Electronic Communications (BEREC) has recently put forth proposals for a single terminating charging mechanism, specifically a shift toward BAK, which it believes will benefit networks in a converged, multiservice, NGN IP-based environment. If implemented in the future, this approach would result in wholesale arrangements similar to those used under Internet peering agreements. In the United States, the National Broadband Plan provides for the Federal Communications Commission (FCC) to adopt a framework for long-term interconnection reform that creates a glide path to eliminate per minute interconnection charges, while providing carriers an opportunity for adequate cost recovery and establishing interim solutions to address arbitrage. Pursuant to this mandate, in 2011 the FCC began consulting on a major overhaul of the interconnection regime in the United States, noting the need to move away from per minute charges, which “are inconsistent with peering and transport arrangements for IP networks, where traffic is not measured in minutes” (United States, FCC 2011, para. 40).

As policy makers consider ways to reform the interconnection regime to enable broadband development, one of the issues to consider is that termination rates have traditionally been a significant revenue source for PSTN-PLMN operators in many countries. This is especially relevant for developing countries in the case of international voice traffic, where incoming calls significantly exceed outgoing calls. Where termination is a major source of revenue, providers may have the incentive and ability to advocate for maintaining wholesale termination arrangements subject to the current switched model (or some variation similar to the current model), notwithstanding the fact that the underlying technical and market drivers will likely have changed. If call termination rates remain high, many PLMN and some PSTN operators may have incentives to choose not to evolve their networks to IP-based interconnection (European Commission 2009, 32). This could have a detrimental impact on the development of converged broadband networks.

However, two factors may favor the transition toward NGNs and IP-based interconnection. First, as networks converge toward NGNs and data services become increasingly dominant, per minute costs for voice services are expected to fall. Second, the ongoing worldwide trend toward regulating termination rates to reflect the underlying incremental costs of termination, especially for PLMN operators, has resulted in a significant reduction in termination rates in many countries. For example, recent regulatory proceedings in countries such as Colombia, Kenya, Mexico, and Nigeria have reduced rates to levels comparable to those prevalent in the EU. As BEREC notes, “The lower the costs per minute and the closer they are to zero, the less difference between CPNP and BAK.” This may also facilitate a transition to IP-based interconnection in many countries.