What do network operators sell to each other?
Since the commercial Internet began in 1995 network operators have explored ways to make internet interconnection work as a market. Yet, a central prerequisite of this market is still ambiguous: the object of exchange. Uta Meier-Hahn traces the difficulty in pinpointing what this market is actually about back to the internet’s architecture and examines solutions proposed by networkers.
Comparisons rarely work. They never fit properly; they almost always lead away from the topic. And the audience‘s attention is lost in unimportant details. Therefore, writers do themselves few favours when they use comparisons as stylistic tools. I once read that in a book about speech writing (Franz, 2015). But there is one exception: the remote comparison. The further away the object of comparison, the better. The remote comparison is intended to show how absurd a situation is. The market at the core of the internet is like a movie without a title.
On 30 April 1995, the internet acquired a screenplay. This established a new setting: the predecessor of the internet, the research network NSFNET was privatised. From this point on, commercial traffic via the backbone of the internet would be fully permitted (Mathew, 2014, p. 42). And everyone was allowed to join in. Of course there were more rules. But what is important about this date is that it marked the beginning of the commercial internet. Hand in hand with the technical infrastructure, the internet was to develop as an infrastructural economic space. Private companies were particularly expected to develop the network infrastructure via the market and establish internet connectivity. But there were few ideas of what exactly the market at the core of the internet would look like and what network operators would trade with each other. One might say that the screenwriters had forgotten the title. But perhaps they had omitted it intentionally.
Markets need objects
The market sociologist Patrick Aspers analyses markets based on five criteria: first, based on what is traded, secondly, based on the social structure of a market, thirdly, based on the interests of actors, fourthly, based on the surrounding culture and, fifthly, based on how the traded things are assigned value (Aspers 2011, p. 82 f.). How network operators do business with each other is interesting from all these viewpoints. For reasons of space I will confine myself here to the first criterion, the what of a market. What rights to services or products do market participants offer each other? Only when this is clear can we speak of a market, according to Aspers. For many markets, this question can easily be answered. The flower, car or currency markets already contain the objects they exchange in their names. In the market at the core of the internet, the situation is different. Here, network operators technically connect their networks and thus jointly create the internet. By establishing internet connectivity, the companies also enter into economic relations. But are these network operators trading something, and if so, what? What is the title of the movie? The picture that emerges from my interviews with 50 network engineers, peering coordinators and industry observers is not a unified one. Most of them find it difficult to define the good or service in this area. Their imaginaries of the objects of the market can be divided into six categories. In practice, these views frequently appear side-by-side or even in opposition to each other.
- Network capacity. By interconnecting their networks, network operators put each other in a position to send traffic to endpoints on the internet or receive it from these endpoints. This capacity is the resource that network operators offer each other. They measure it in volume per time, that is, in bits per second. Networkers determine the quality of network capacity as an object based on criteria such as packet loss, latency or jitter.
- Routes. When they are interconnecting, two networks communicate to each other what destinations on the internet they can reach for the now connected network via what pathways between two points in the network. Two types of arrangements can be distinguished here: in so-called peering, the interconnection only involves networks opening routes to endpoints in their own network and their customer’s networks, but not to endpoints for which they must pay them. Destinations in this case refer to IP addresses. The unit of commerce in this sense is the route, that is the precise pathway. In cases of transit, the transit provider offers the transit taker more, namely the opportunity to exchange data with all endpoints on the internet. Networkers assess the quality of routes based on the length of the pathway – the shorter, the better. Seldom-used routes are worth more than those that are easily available.
- Usage rights for the physical connection. From this perspective, the product is made up of the individual components that require an investment of effort or money by a network operator. That may include network resources such as routers, switches or fibreglass, as well as personnel, licenses or electricity. In this understanding, the market service rests on the fact that network operators grant each other the right to use physical and logistical infrastructures.
- The exchange of external resources. In the case of peering – recall that here, only the peering partner’s own and their customer’s routes are shared – some networkers demand that their interconnection be understood as an exchange relationship. Money may also flow, but it does not have to. A commonly used phrase describing the exchange is content for eyeballs. Networks on one side provide direct access to content; networks on the other side provide access to end-users. An internet provider commented: “I am selling access to credit cards.” But content and access to end-users are not the only resources that network operators exchange. It is said that banks – yes, even banks and insurance companies now operate networks! – offer other network operators financial services for access to their internet users. In both variants, the technical parameters of the interconnection itself slip into the background of the exchange relationship. Instead, the actors bring external resources that they produce or control into the negotiation.
- A connection-oriented relationship. This notion relies on a feature of the technical architecture of the internet. This involves neither an item for trade nor a service model. Network operators exchange data using the Internet Protocol (IP). The IP is a so-called stateless packet switching protocol. IP divides contents in small packets for transport. These packets can be theoretically routed to their destination via different pathways and through many interconnected networks on the internet. At the destination, an application reassembles them. Unlike in the postal system or when airlines work together, network operators do not transfer entities to each other, such as parcels, or rights to seats on the plane; they only transmit disjointed fragments. The network itself thus cannot register any transactions. Instead, there is only one persistent connection between two networks, that is, a relationship. The connection itself and the understanding of roles that emerges as a result of this relationship becomes the subject of commercial negotiations. Even without a specific object of exchange, situations may arise where a network ultimately pays for the interconnection. This may be the case even when both parties achieve cost savings in the process, because they exclude third parties from their business by creating a direct peering connection between each other. Such scenarios are called paid peering.
This role distribution leads to a kind of evaluation poker. Every network operator seeks to convince the other that the other party is benefitting more from the interconnection and should therefore pay. One interviewee likened this game of nerves to an encounter between two animals in the jungle at night: both only see the reflecting eyes of the other. To avoid a fight, one animal must give up and come out of the situation as the weaker party. In this framing, the negotiation between network operators does not lead to a disclosure or an actual equalisation of power. Instead, it remains a symbolic power struggle, which nonetheless assigns the two network operators roles in the economic relationship – they become customers and vendors, or they become peers.
- No market at the core of the internet. Many network operators state that the direct interconnection of networks during peering takes place in a market environment. However, they consider these markets as something that does not relate to the interconnection itself. They do not identify themselves as participants in what others call the interconnection market (Zarnekow, Wulf & of Bornstaedt, 2013). They see themselves primarily as participants in markets that are external to the interconnection in which they act as providers. This is the case when a content delivery network announces that its product will speed up content on the internet, when a social network indicates it will deliver 100 per cent of the product online, or when one of the world‘s largest internet providers claims that peering improves its actual network products. In all of these cases, the network operators see the value of the peering connections in the fact that they facilitate business in their end-customer markets (‘it’s an enabler’). They understand the direct interconnection with other networks as a precondition for their own value creation, but they do not recognise that this value creates a market in itself. Some of these network operators have publicly proclaimed that internet providers should have end users, and not other operators, cover their costs.
Variety of framings leads to uncertainty
What can this variety of answers say about how network operators do business with each other? First, it is evident that the categories significantly differ. The services in this market appear to be nebulous or even contested. The diversity of views network operators on the object traded initially suggests uncertainty in the market rather than stability.
The first three categories are compatible with the market principle: network capacity, routes and the physical connection. What they share is that they build on relatively strong levels of formalisation. This allows network operators to compare offers. Bits per second, routes or the cost of hardware can be approximately measured and provide starting points for an assessment. However, these categories are not globally standardised among network operators. Even the perhaps most standardised product, IP transit, can have different characteristics. Perhaps that is one reason why there are no comparison sites such as Amazon.com for connectivity services. Currently, network operators are mainly finding out who is offering transit where and for what price, by word-of-mouth. More consensus on the what of the market would obviously increase transparency, facilitate evaluations and thus increase competition among network operators. This is not favoured by everyone, as shown by the controversy surrounding the initiative by two network engineers, who publicly compare the offers of internet exchange points with each other. 
The last three categories (the exchange of external resources, a connection-oriented relationship and no market) cast doubt on the concept of an interconnection market itself. The initial question about the what of the market that is assumed to exist at the core of the internet brings more forms of economic coordination to light, namely bartering and (social) network coordination. These two forms of coordination operate without competition. For the bartering concept, it is unclear how widespread it is and under what circumstances network operators opt for it. In contrast, network coordination can be observed frequently. It is based on reciprocity. Network operators mutually relate to one another and cooperate. They have to do this to make the interconnection technically possible. As the example of the jungle shows, cooperation does not do away with hierarchical constellations and struggles for power (cf. Mützel 2008 p. 191 f.). The last category, namely that a market at the core of the Internet is unacceptable, certainly has a political character. A person who identifies him or herself as a market participant only outside of the context of interconnection denies the legitimacy of those who want to develop markets around peering arrangements. So-called peering wars are therefore disputes over whether economic relations between network operators should be based on reciprocity or on the market principle.
Even more than 20 years after the commercialisation of the internet, the application of the market principle at the core of the internet is still contested. The internet’s architecture makes the object of exchange seem constructed, which induces uncertainty. Nevertheless, network operators obviously do trade with each other. But if one asks on what the stability at the core of the internet is based, you have to look further than the market and its object. Because the participants have not yet agreed on a title for the film.
This post is part of a weekly series of articles by doctoral canditates of the Alexander von Humboldt Institute for Internet and Society. It does not necessarily represent the view of the Institute itself. For more information about the topics of these articles and asssociated research projects, please contact email@example.com.
- Aspers, P. (2011). Markets. Cambridge ; Malden, MA: Polity.
- Franz, M. (2015). reden schreiben wirken: und ganz nebenbei ein besserer Mensch werden. Correctiv.
- Mathew, A. J. (2014). Where in the World is the Internet? Locating Political Power in Internet Infrastructure. Doctoral dissertation. Retrieved from http://www.ischool.berkeley.edu/files/ashwin-dissertation.pdf
- Mützel, S. (2008). Netzwerkperspektiven in der Wirtschaftssoziologie. In A. Maurer (Ed.), Handbuch der Wirtschaftssoziologie (pp. 185-206). Springer.
- Zarnekow, R., Wulf, J., & von Bornstaedt, F. (2013). Internetwirtschaft – Das Geschäft des Datentransports im Internet. Berlin, Heidelberg: Springer Gabler. doi:10.1007/978-3-642-36687-1, pp.60-95
 This table lists the largest European internet exchange points, including the prices that network operators must pay for 10 und 100 Gigabit Ethernet connections there – http://goo.gl/4D4Ur3[/footnotes]
This post represents the view of the author and does not necessarily represent the view of the institute itself. For more information about the topics of these articles and associated research projects, please contact firstname.lastname@example.org.