Why spectrum is not property - the case for an entirely new regime of wireless communications policy
David P. Reed (http://www.reed.com/dpr/locus/contactinfo/)
February 27, 2001 DRAFT

What we now know about the physics and architecture of RF communications contradicts the "property" model of spectrum.

Yet the entire tradition and practice of managing wireless communications technologies in the broadcast, communications, and telephony industries is firmly based on a legal "metaphor" that equates spectrum allocations with rights in physical property, such as land use rights. Since uncoordinated use of spectrum can lead to problems of interference, coordination of spectrum usage is known to be necessary. The "metaphor" of property rights has been used to structure that coordination by subdividing available capacity among users so that interference is avoided.  This subdivision assumes that the available capacity that can be used with a given technology is fixed, so that the goal is to apportion the fixed capacity among the uses with the goal of optimizing overall societal benefit.

The Problem

One example should suffice to show that this metaphor no longer works. In a nutshell, the problem is this. We now know that we can arrange for the capacity of a fixed amount of spectrum in a fixed volume of space to increase as the number of users increases.

Here's the example. Two independent research projects (Tim Shepard in his MIT Ph.D. thesis, and Gupta and Kumar of University of Illinois, Urbana-Champaign in a recent paper) have shown that the capacity of "spectrum" can be managed so that it increases with the number of users. The key idea is to organize the spectrum users into a cooperative network rather than an uncoordinated set of point-to-point channels. As the number of users N increases, the capacity can grow as N1/2 (the square root of N) in the architectures they suggest. Thus, the more users that make up the network, the more capacity the network can carry.

Further, I have good technical reasons to believe that future cooperative wireless network architectures can achieve a capacity that scales proportional to N. If that does turn out to be true, then adding each new user creates sufficient capacity to support that additional user. There's no strong proof that O(N) is the limit among all possible architectures.

In addition, the flexibility (quantified by the set of options created within a particular wireless network architecture - its "optionality") can grow more steeply than N - e.g. N2 for Metcalfe's Law that quantifies the growth of pairwise transaction options, and 2N according to my own (Reed's) law that describes the growth of options in group-forming networks. (See http://www.reed.com/gfn/ for more on scaling laws for option value in networks.) For many applications, flexibility of this sort is as important as capacity in defining the economic value for users, i.e. utility.

What a new regime must do

Since each of these scaling laws show that cooperative wireless networks can create more utility as the number of network users increase, then "spectrum" capacity for communications does not behave like any ordinary kind of property. We create markets to manage real property and other goods because that is an effective way to avoid "The tragedy of the commons" so well-described by Garrett Hardin in his essay by the same name.

But when the commons' capacity can increase with the number of users, we clearly need a different regime to allocate that capacity among users.

Such a regime must encourage cooperation among users, because non-cooperation can prevent the creation of additional capacity.

We don't know today what the "best" cooperative architecture will be. Surely Shepard's design is only the first among many that will be discovered by researchers and innovators.

So any new regime must also encourage innovation in new architectures, and foster creative competition among various architectures to search for the most effective combination of architectures and policy.

So what should we do?

I think now is the time to look backwards to the early days of the Internet for inspiration. When the Internet Protocol (now called IPv4) was created, we did not know what the best technology would be for building networks. Today we are using technology that was never contemplated in the late 1970's to transport Internet messages, and yet that protocol continues to be the core of the Internet. The key idea in creating IPv4 was decoupling the users of the Internet from the properties of the technologies inside the many transport networks that make up the Internet.

We need a regime that allows RF networks to interoperate and cooperate in use of "spectrum" in an open and experimental way, just as the Internet did.

The design of this regime, which spans physics, architecture, economics, and policy, is a worthy goal. It is more than a "research project" and more than a "spectrum allocation policy", because it needs to consider all four areas at the same time.

While the active work on such areas as "space-division multiplexing", OFDMA, "mobile ad hoc networks" and "software defined radio" contribute to the technology that may enable a more interesting future, these technologies by themselves are not sufficient, and are not even strictly necessary. In the early Internet, for example, the power did not derive from the invention of optical fiber or the Ethernet - instead it derived from an architecture that could absorb unanticipated innovation into its fabric without a speedbump. The Internet architecture was fundamentally a "cooperative" network, which created huge incentives for cooperation among it many builders and users. We needed the Internet architecture to create a space in which the last 25 years of innovation could create a new communications regime that is quite different from the regime that grew up under the primary control of the telephone company.

A call to action

We need to see the "spectrum" in a new way today. There is an opportunity for much better spectrum use that is based on cooperative networking rather than a property model. Without a doubt, the opportunity is huge. It's time to get started.

For references, see http://www.reed.com/OpenSpectrum/

Copyright © 2001 David P. Reed. Online citation by URL is explicitly permitted and encouraged. For other uses, please request permission of author.