Some thoughts about IP interconnection – blog post number one

Silicon Flatirons, a respected technology policy think tank housed at the University of Colorado, recently held a workshop focused on IP interconnection issues. I participated in the work shop. The ground rules of these forums prohibits me from discussing specific comments made by participants or from discussing in any detail what transpired. But the session did inspire me to write a series of blogs offering my own thoughts about three key issues regarding IP interconnection – the technological underpinnings of IP networks; some key economic aspects related to IP interconnection; and innovation in IP networks. This first blog looks at the technology issues.

The old telephone network (also fondly referred to as POTS, Plain Old Telephone System or PSTN) was an amazing thing for its era. Imagine what it must have felt like to be able to dial a few digits and within seconds to be able to speak to someone halfway around the world, in an era where a physical letter took many days to get delivered, and email and WWW were typos, not services. The PSTN was “centrally” managed – by the telephone companies. "Dumb" peripherals - your basic black phone - connected to a "smart" network, one in which the network functionality was simple in one sense (i.e., routing calls) but complex in the sense that the calls had to be measured and billed. The network offered one service: voice. The network was aware of who initiated the transaction (in the sense that it “knew” physically where the line terminated and who to bill for it), who “terminated” the transaction, how long the transaction lasted, and what carriers were involved in supporting the transaction. This provided a way to allocate the costs of the operation of the network to those who used the network. In general, there was one call initiator and one call recipient. Life was simple.

In this model, physical interconnection was much more immediate and closer to the source of the switching function. Since a physical circuit had to be "built" end to end from the source to the destination in order to connect and compete, a competing carrier (CLEC) needed to have access to the lines managed by the incumbent carrier (ILEC) in a central office, because these lines were connected physically to a person or business and associated with a specific phone number.

The architecture of the Internet turned all of this on its head. It is not centralized. There no longer are only a few networks managed by dominant players who have a central and controlling role (the former monopoly telephone companies) from protocol specifications to network build outs. Now, there are thousands of networks, many not operated by the traditional “carriers” but instead by web site providers, content providers, private companies and the like. The protocols are open and managed more collegially. What used to be an easily measurable transaction - call duration and distance - is now no longer easy to measure. A transaction is now an attribute of the application that is run on the end systems, and the network is generally not aware of these transactions. All the network sees is packets, and each packet provides limited visibility to the network about what it is carrying.

This is even more true in the case of encrypted traffic, where data is simply invisible and impenetrable to everyone but the source and destination. Now that the “transaction” is not visible to the network operator, this implies that any effort for a network to use a transactional service tariff model becomes an exercise in frustration. There are no such network-visible interactions that could be used to create a transactional service model.

Interconnection in a “switching center” is not necessary or vital any longer because physical access to individual lines is not needed as was true with telecommunications and telephony in the past. Now, packets can be routed to a physical location virtually, and the physical interconnection can happen in many ways, in many different forms. Just because an interconnection agreement is not reached with one player does not mean that it is not possible to get your traffic where you want it. There are, in other words, many ways to skin the cat. That was not true in the past. John Gilmore's quote - "The Net interprets censorship as damage and routes around it." - applies whether one is referring to censorship by a state or lack of an interconnection agreement between two players. There are many pathways to ensure traffic gets to where it is going. This is a part of how the system was designed and it has proven itself to be highly adaptable and capable of evolving.

There is no one "Internet" to govern, and the coalition of networks is not governed by any one entity. If anything, it is governed by everyone. While the peering and IP interconnection (i.e., transit agreements and the like) process may not be a multistakeholder led activity, it is a form of governance because it is about making the Internet work. Agreements are reached among networks all of the time that keep Internet operations moving. All governance is not done via multistakeholder organizations, but it happens in a number of ways and generally without direct government involvement. IP interconnection is an example of this in action. If this were not the case, the Internet as we know it would look very different.

Thanks to my colleague, Sanjay Udani, for some useful edits and guidance.