Friday, February 28, 2003

We didn�t create the automobile by lashing a carriage to a mechanical horse but we were able to repurpose the roads designed for horses by paving them to create a smooth surface. The Internet isn�t just an upgrade to the phone network. It needs its own path. The existing copper infrastructure is a valuable resource that can be used as a native medium for Internet connectivity. We must take advantage of the opportunity to provide universal connectivity very quickly at a low cost, we get vastly improved telephony as a free bonus.

It's well known that there is a lot of "dark fiber". High capacity networks run over glass fibers and you light them with a laser to transmit data. When burying cables the cost of adding extra fiber is low. These fibers are kept dark until needed.

What is less well known is that we have a similar situation with our copper-based or traditional infrastructure. We generally "light" it with the low frequency based-band signals that we've been using for telephony for the last century.

If we light it with radio signals (as we do with DSL), the capacity goes from a few kilobits per second to megabits. Since we dedicate a pair per household we are leaving the wire completely empty almost all of the time. If you use the telephone for two hours a day at 32kbps (to be generous), then you are sending only 3kbps on the average on a wire that has a capacity of a few megabits per second. This means we are using only .1% of capacity but I'm generous so I'll say 1%.

According to Dave Burstein, the cost is a few hundred dollars a port. Since it can be done incrementally this is a very attractive way to upgrade the infrastructure. While I understand the desire to run new fiber -- it does provide far more capacity -- it is hard to justify throwing away an infrastructure that can deliver connectivity now and where the demand is. While fiber might be useful when we want to deliver 100 megabits to each home, we must not have a static goal. We must be able to buy capacity as we need it. If we can't get enough capacity over copper, then we will deploy fiber. But not just 100 megabits, we will have a market process that can deliver as much capacity as we demand. As we've been with PCs, 100 megabits may seem like a lot now but in a few years we'll view it as too slow to even consider.

Lighting our copper infrastructure can give us the benefits of connectivity now (and it could've given us the benefits five years ago). The benefits are even greater if we treat the bundle of copper as a whole rather than reserving a pair of wires for each subscriber. We would then have a multi-gigabit pipe that is tolerant of failures. It is easy to extend the service by mixing copper and fiber segments since they are just transporting bits.

At the central office we also get very large savings since we needn't do any telephony processing at all. The voice calls are just low speed data streams that add little additional traffic. Phone calls only incur a charge when bridged to the legacy phone network (The PSTN or Public Switched Telephone Network) and that will become the rare exception. Even then, there would only be local calls and, in the United States, such calls generally incur no additional charges.

Given that I can buy a full computer that acts as a network router and provides NAT (Network Address Translation) and other services for $20 and a PC modem is often implemented entirely in software, the actual cost of lighting the copper to capacity should be very low. Resilience is far less expensive than assuring that no component will fail.

Connectivity means that we are not longer dependent upon a very slow and cumbersome process for delivering new services.

E911 is a great example. The E is necessary to distinguish it from 9/11. In fact 9/11 only heightens the need for improvement. While we can convince ourselves that E911 works, it really doesn't work very well. A single cut wire makes it fail. And it only identifies the phone used and that phone might not even be where it is supposed to be. Any "improvement' takes decades. While E911 is a great improvement over nothing, we can do far better.

An IP based E911 could evolve very rapidly. At very least a heartbeat would detect the loss of connectivity. One can hire additional observers, such as security firms, to monitor the messages. Instead of just a single phone, the system can allow for additional sensors and signaling. Patient monitoring would be an integrated service instead of a special case. Once we can have additional observers we can also use the system to provide a very detailed map when there is a disaster such as a flood. If the wires are out we can use wireless connectivity instead even if it is at a lower speed. This is all very simple and low-tech once you have connectivity.

Of course, there are many additional services that provide immediate economic value. These services are currently stymied by the ancient telephony paradigm which is built upon circuits that require exclusive use of a particular pair of copper wires while providing connectivity between only two end points at a time. The Internet shares these resources and connects everything to everything.

Just as we don't treat the car as a horseless carriage, we should stop thinking of our copper infrastructure as the telephone network. It's just a dimly lit neighborhood off the Internet waiting for the light to shine.