I gave a talk yesterday at Northwestern called A DNS in the Air [PDF]. My idea is that, in order to scale, the emerging wireless Internet needs something analogous to the domain name system (DNS)—the infrastructure that allows you to reach sites across the Net.

Billions of mobile phones, and even more billions of connected sensors and other wireless devices will completely overwhelm our current spectrum management regime. AT&T Wireless estimates we will need between 250 and 600 TIMES the current wireless capacity in 2018, less than a decade from now. Wireless has to change in order to scale capacity, just as the Internet changed data networking. The only way to do so is to get beyond the “command and control” regime of spectrum management, which leaves the vast majority of frequencies (95% on average in US cities) sitting idle most of the time. We need opportunistic means of sharing spectrum, including dedicated unlicensed bands as with WiFi, but also allow for overlays and real-time markets.

The DNS solves the problem of coordinating addressing across a global inter-network. It does so through an ingenious design, heavy in distributed redundancy, and by separating out policy issues from the mechanics of routing. We need something similar for wireless. And we may be in the process of building it.

The FCC decided last November to authorized unlicensed “TV Band Devices” in the “white spaces” between and around television broadcast channels. Google, Microsoft, the New America Foundation, and others did great work to persuade the FCC to go forward with this idea. One requirement in the FCC order is a geolocation database. White space devices must check this database to identify where they can transmit, and which frequencies should be avoided to protect licensed users (TV stations and wireless microphones primarily).

The White Space Database can become the DNS of the wireless Internet… if it’s done right. The database needs to be expansive enough to include all spectrum, not just the TV white spaces. That will require a real spectrum inventory, something Congress and the FCC finally seem willing to encourage. The database needs to be distributed to avoid single points of failure and excessive concentration of control. And we needs to separate out and think through the operational, policy, and governance issues. If this happens ahead of time, we can avoid some of the problems that have bedevilled ICANN, the oversight body for the DNS.

I’m writing a paper that will expand on this idea in more detail.