The Core and the Edge

Think of this as Volume 18, Number 12 of the newsletter I have written weekly since March, 1997. Enjoy.

The IN in the word Internet stands for interactive. (Right, Sir Tim Berners-Lee, whose 1989 paper set the terms for today's World Wide Web.)

The assumption of its designers was that the Internet would be a two-way network, that as many bits would flow in one direction as the other. Peering, the original financial agreement underlying today's Internet, is based on this assumption.

The Internet, in other words, was not designed to be TV. It was not designed to be a passive medium, as something you watch, or read, or listen to. We already had such media. The Internet was designed to be something you interacted with.

That's not how things are working out in our time. The Internet is becoming TV, is becoming something you watch, or read, or listen to. The bits are mostly flowing in one direction, from the core to the edge of the network. Most of the technical standards underlying consumer Internet traffic, like DSL and DOCSIS, are designed to let bits flow faster downstream than upstream, on the assumption that more bits would do that.

There are also different cost structures at the core of the Internet and on the edge. When you improve something in the core, that improvement goes to everyone at once. It's relatively cheap. Changing the edge means changing all client devices at the edge, and it costs more. Improving a fiber link can be as simple as changing the diodes at each end of it. Improving local networks may mean rolling trucks, and digging up streets.

The core of the Internet can take advantage of Moore's Law. The edge does not.

This is true for non-Internet networks as well. Payment networks are ready to take EMV or chip cards. They were improved by programmers from central points. Issuing banks, however, have to send out over 1 billion chip cards, at a price 10 times greater than mag-stripe cards, to cut the $5 billion in fraud on that $4.4 trillion market by, maybe, half. And merchants have to install EMV terminals, and cash registers. They also have to train people in new check-out procedures. It's expensive.

All things being equal, all things are not equal on the Internet. Not as they relate to costs. And not as it relates to benefits, either. Core services like Facebook and Google are much better investments than edge services like AT&T. Core services are what people want from the Internet. Edge services are mere means to that end.

As a result of all this, improvements in the Internet's core have outstripped those at its edge. Core services complain that there isn't enough speed at the edge, that consumers can't get everything they want to deliver, that U.S. broadband is falling behind that in other countries. They're right.

When the edge services try to fight back against this economic reality, they are accused of violating “net neutrality,” the principle that people should be able to access any legal service they care to, without discrimination. They are doing just that, because they can't make a single agreement with the core, only with specific interests inside that core.

Beyond trying to sign new agreements with core services, edge services are also fighting back by buying “content” services that are delivered from the core, but desired at the edge. When BT won the rights to broadcast Premier League football matches, and then the Champions League, that was an example of the edge controlling the core. When Comcast and other cable companies began their “TV Everywhere” effort, in which their subscribers could access cable programming from other devices, this was also an example of the edge seeking to control the core.

A few generations ago, regulators fought this kind of vertical integration. The fall of the Hollywood studios is directly attributed to Washington's demand that theaters be divested by studios in the early 1950s. Without that integration, the studios withered. When they came back a generation later, as TV networks, and demanded the right to own their own programming, Washington gave in.

When NBC broadcasts a show, or a game, Comcast writes a check to itself. When this programming is offered through Comcast as an ISP, Comcast writes a check to itself. Vertical integration is very profitable to the edge, and it's possible because the capital costs of the edge make programming assets vulnerable to purchase. It also limits consumer choice. It's a direct assault on net neutrality.

The fact is, neither side is being honest about the issue. Core services aren't admitting to their financial advantages, over the edge. Edge services aren't acknowledging the inherent conflict of interest in owning content.

Consumers are being whipsawed by rhetoric from both sides, and policy makers in Washington are suffering from the same thing.

But we don't have to be. It is time to acknowledge that there are cost differences between the edge and the core. It is time to build business models acknowledging this cost difference, and enabling improvements in edge performance the core needs in order to grow. But it's also time for edge services to stop lying about what they're doing. They are violating “net neutrality.” They are trying to control Internet consumer behavior through their control of vital infrastructure.

The edge deserves to be paid. But the Internet deserves to be free.