Last week, a lot of attention was focused on a dispute between Comcast and Level 3. The facts of the dispute seem to be reflected in each party's ex parte letter to the FCC. Here is Level 3's version, and here is Comcast's version.
What seems not to be in dispute is that Level 3 acquired additional demand for data destined for delivery to Comcast's broadband customers as a result of Level 3's acquisition of content delivery business from Netflix. Comcast asserts that Level 3's increased demand for capacity would impose additional, unplanned capital costs on Comcast in order to accommodate Level 3's increased capacity demands. When Comcast asked to be paid for these additional ports, Level 3 contended that Comcast was violating the Commission's principles of net neutrality by price discriminating against traffic that Comcast knew was originated by a rival video content provider. Comcast responded that Level 3's demand for additional, asymmetric capacity was a violation of the parties' "peering agreement."
While neither Comcast's nor Level 3's description of their disagreement seems to accurately characterize what is happening between the parties, it is not surprising that each party's characterization of the dispute uses its own self-serving nomenclature to cast itself in the best light. What is notable, though, is that both parties' choice of nomenclature centers around the customary means of commercial Internet traffic settlement--among large transiting carriers--known as "peering."
Peering is essentially another name for "bartering." For "barter" to be the equilibrium method of traffic exchange, it is necessary that the parties be involved in the same business (Internet traffic transit and termination), offer each other roughly the same potential value (which neither party has on its own), and--over some temporal period--provide each other the same real value. In other words, over time each peer must offer a combination of scope and capacity, which despite daily fluctuations, ends up providing the other with "fair" value without the need for additional consideration.
So if the amount of one carrier's traffic that it sends onto another carrier's network exceeds a certain allowable asymmetry (say 2:1) over time, then that carrier is "using" more of the other carrier's capacity and must become a "customer." "Customers" of Internet backbone service purchase transmission carriage, or "transit", on the network of the backbone provider they wish to purchase capacity. It is worth noting that in almost every case that there is a peering relationship, even peers (from time to time) will have to purchase transit on another operator's network to preserve their own peering relationships. So even "peering" is not pure barter. This is a rough simplification of how the "tier 1" Internet backbone "peering" system developed and works, but, to be clear, even "peering" isn't free.
Even given this limited description, it seems that that both Comcast and Level 3 have publicly characterized their dispute inaccurately. Comcast--for purposes of Internet traffic carriage--is (for most routes) not Level 3's "peer." Level 3 has an extensive worldwide network capable of addressing well over 100,000 traffic destinations. Comcast does not. However, neither is Level 3 (as a content delivery network, or "CDN") a "peer" of Comcast (as an Internet service provider, or "ISP"). This is what makes the dispute interesting, and lets the elephant into the room. How much longer will the "peering" method of capacity trading continue to be considered the "norm?" One thing seems certain, and that is that barter (peering) rarely remains the default equilibrium in any market over time.
While it is always tricky to try to compare and draw inferences from one industry to another, let's be nutty enough to try. One example comes from an antitrust case I have previously discussed on this blog: Aspen Skiing. Let's focus just on the facts of that case. There were three mountains in Aspen that skiers found attractive--they wanted to ski multiple mountains on their trip to Aspen, but they found it inconvenient to purchase multi-day lift tickets from multiple operators. To better serve their customers, the two operators of the three mountains decided to come up with an "All Aspen" ticket for one fixed price. In other words, as competitors they cooperated to bring a consumer-friendly offering to market. This "peering" system worked . . . until it didn't--which gave rise to the antitrust case. And, while the winner of the antitrust case got monetary damages, the "All Aspen" peering equilibrium never returned.
The history of bank ATM networks are a similar story. Banks wanted customers to use machines rather than tellers, in order to lower their costs. To make using machines attractive, banks invited many local banks into their "networks" so that customers would see the value of being able to get cash from any machine in the network--regardless of bank owner. The banks operated the networks on a non-profit basis, with each paying an annual fee to share network costs and a nominal fee when their customers used another network member's ATM (they also received a nominal fee when another bank's customers used their ATM). Notably, no customers were ever charged in the early days of ATM networks, rather the banks acted as, more or less, a barter cooperative. As we all know now, this system, too, worked . . . until it didn't. Banks with a greater number of ATMs felt they were offering a greater service, and began charging non-customers to use their ATMs. Thus, from the perspective of end-users, one-time "peers" became "customers."
As carriers of information and telecommunications--regardless of technology--begin to resist "traditional" defined roles, they seem inexorably on course to resist "peering" as a method of traffic exchange. Network owners want to be paid for the use of their networks--and they largely function that way now. The amount of traffic subject to "peering" as a means of settlement seems to be bound for decline--as the amount of "same traffic" (like world wide web traffic) declines.
This is the truly interesting issue highlighted by the Comcast-Level 3 dispute. There are a number of questions that must be answered. If Comcast's broadband users are "pulling" capacity onto Comcast's network, should the heaviest users of services like Netflix pay the costs of accepting that additional demand? Does overall demand capacity really matter as much as where in the terminating network the capacity is needed? Who controls cost causation? Do the "traffic destinations" have an incentive to share network capacity planning with networks seeking to deliver traffic to their end users, such that the most cost efficient outcome for both networks and their customers is achieved?
These are the questions that must be resolved in the new world of non-peers exchanging capacity-based traffic. This isn't a bad thing, and the industry--diverse as it is becoming--may have the incentives to develop efficient answers to these problems. However, shedding "tears for peers" and trying to shoehorn disputes into frameworks that no longer fit is not going to produce a system of settlements that is satisfactory to all market participants, and accommodative to future traffic generators, transit providers, and traffic terminators.