The Economics of Internet Interconnection Agreements
EN: This chapter also deals with interconnection agreements, examining the economic, technical, and policy perspectives of different models. Again, this stems from a short-term problem of backbone capacity that has, for the present, been solved, but it is worth considering from a theoretical perspective.
Incentives for Interconnection
The nature of the Internet is such that at least two service providers are involved in each request: the ISP that provides access to the customer who sends the request, and the ISP that provides hosting to the customer whose server receives the request by sending data back the other way.
The primary incentive for interconnection is an economic one: even without legislative requirement, an ISP is inclined to accept some traffic from other networks onto it s own in exchange for its ability to do the same - the only alternative being for each provider to build a redundant network to every possible destination.
This alternative is not merely theoretical - it describes a private network, which is a technology that predates the Internet and, in some instances, is still utilized where connectivity and confidentiality is critical.
Interconnection Architectures
The resale model is, at heart, a kind of interconnection agreement, in which one provider who does not have a network that covers a given area purchases network capacity from one who does - but instead of giving its vendor access to its own network, it provides payment.
However, there are instances in which network providers exchange traffic for traffic without money changing hands. Four interconnection architectures are described:
A peer-to-peer bilateral model involves two providers of similar size, with similar capacity, who enter into an explicit agreement to share network traffic among their network nodes. Peer status, especially in terms of similar capacity, is necessary to ensure that each provider has the capability of upholding its end of the agreement.
A hierarchical bilateral agreement is also governed by an explicit two-party agreement, but ti exists when the two players are not equals in terms of capacity, coverage, or market power, such that the agreement is made conditional to even the balance.
The third-party administrator model is when more than two networks are interchanging packets and the administration of the interconnection is done by a firm who does not operate either network. A more concrete example deals with internet advertising, where multiple publishers allow a third-party service to control the display of ads by certain agreed-upon rules.
A cooperative agreement is similar to the above, but the administration is controlled by individuals appointed by those who are participants in the network based on decisions made by a committee. The author admits this may be difficult to implement among competing commercial interests.
Entering into any such agreement requires trust, such that each party abdicates opportunism in exchange for others doing the same. Should one party violate the agreement to the detriment to another, the agreement will likely fail and the interconnection of networks will desist.