TCS Daily


Moore's Bailiff

By Arnold Kling - July 17, 2002 12:00 AM

At Longbets, people make long term forecasts, and they put their money where their mouths are. Andy Chapman, founder of Narad Networks, which offers broadband software for cable companies, predicts that at least one local telephone company will require a government bailout to avoid bankruptcy by 2007. I think he is somewhat aggressive about the date, but in terms of economics, he makes a good forecast.

People say that economic forecasting is difficult. I think that forecasting is easy, as long as you stick to forecasts that are based on Moore's Law. Moore's Law is the most powerful economic discovery since compound interest. In fact, Moore's Law is much like compound interest, in that it is a nonlinear process with large cumulative effects.

As I will explain below, I believe that Moore's Law means that the phone companies will lose the broadband battle. Chapman's view is that once they start to lose, they will collapse.

    Their capital structure insures that...even modest market share losses...will cause them to lose creditworthiness and, thus, access to debt capital.


How Moore's Law Created the Internet

The Internet can be viewed in part as a technology that competes with telephony. The Internet uses packet switching, which means that messages are broken up into packets. To get a message from point A to point B, several packets must be sent, each one requiring connections to be opened and closed.

In contrast, telephony uses circuit switching. When you make a connection (complete the circuit), you keep the connection open until the message (phone call) is complete.

As economists Hal R. Varian and Jeffrey K. MacKie-Mason pointed out in a brilliant early analysis of the Internet, the viability of packet switching depends on the relative cost of lines vs. switches.

  • ·A packet-switched network (the Internet) makes efficient use of lines, but it requires lots of switching. The same line can carry multiple packets, but each packet requires switching operations to enable it to reach its destination.

  • ·A circuit-switched network makes efficient use of switching, but it requires lots of lines. A phone call only requires that a connection be made once, but each phone call uses line capacity, even when people are not talking.


The cost of using switches to open and close connections depends on the cost of computation. Because of Moore's Law, this cost has plummeted. As a result, the cost of operating a packet-switched network has fallen much faster than the cost of operating a circuit-switched network. Without Moore's Law, the Internet might have remained a Defense Department curiosity, because no one else could afford to use it. With Moore's Law, the Internet protocols are poised to take over all communications.

Moore's Law, Wireless Networking, and the Last Mile

Moore's Law is also starting to have an impact on the use of the wireless spectrum. The traditional model allocates spectrum bands to particular users, such as television stations. The new model, advocated by Frankston, Reed, and Friends among others, is to let many users share spectrum, and let the receivers sort out the messages.

The old model of spectrum allocation allows crude receivers to work, by keeping out signals that the receiver might not understand -- what used to be called "interference." The new model says that receivers with sophisticated processors do not need to be protected from irrelevant signals. It says that "interference is in the ear of the receiver." With sophisticated receivers, people can send signals at the same time in the same frequency band and still be understood.

Once again, we have an old model that is designed to economize on computer processors while making inefficient use of transmission capacity. We have a new model that allows a given transmission capacity (in this case, spectrum) to carry more information, but requires more computer processing. Guess which model benefits from Moore's Law?

As Timothy J. Shepard puts it, 50 years ago, no one could have designed a workable system for shared spectrum. 25 years ago, a few engineers might have been able to come up with such a system. Today, many engineers could do so. The possibility has been opened up by Moore's Law.

The Long-Term Architecture

The way I see it, Moore's Law ultimately will favor shared-spectrum wireless as the solution for last mile connectivity. Today, I am typing this out on my porch, using a laptop that connects wirelessly to a router in my basement, which in turn connects to the local phone company by DSL. My prediction is that eventually I will skip the DSL part, and instead my wireless connection will go to a local wireless network of some sort, and then ultimately to a transmitter on the Internet backbone. The communication network will have a fiber skeleton and a wireless skin. Telephone land lines will be superfluous.

None of this is going to happen next month or next year. I am not sure it will even start to have an impact soon enough to enable Andy Chapman to win his bet. But as the bailiff of Moore's Law, I think that eventually the local phone companies are going to be run out of town.

 

Categories:
|

TCS Daily Archives