Imagine this telephone company advertisement: "DSL – all the network connectivity you'll ever need." It's a joke, right?

Technologically speaking, fiber wins. It is the end game. But fiber access has two problems: big-capital economics and telco-shaped policy. Fiber is the direct route to shipping the most bits per second per dollar, but initial construction costs present huge barriers. The current capital crunch may have brought new fiber construction to a standstill.

And there are bigger problems – fiber cables are enough like copper twisted-pair that telephone companies can use 100 years of legal and regulatory know-how to exclude and impede competition based on routing and stringing cables

Meanwhile, the need for high bit-rate connections remains. The longer we have to wait to get fiber, the more attractive less optimal solutions, including wireless ones, become.

At present, unlicensed wireless access – such as the Institute of Electrical and Electronics Engineers Inc. (IEEE)'s 802.11b, also known as WiFi – delivers speeds measured in megabits. While 802.11b is slower than fiber and much more complicated, it nonetheless provides easier access than fiber because: (a) it is unlicensed, therefore difficult for incumbent communications companies to control; and (b) you can buy it off-the-shelf, plug it in, and use it today.

But 802.11b may be short-lived, worries wireless pioneer Dewayne Hendricks, who heads the Dandin Group Ltd. Hendricks is concerned that as 802.11b gets popular, its very popularity will make it harder to use. The 2.5GHz band could become so crowded that nobody will want to go there. Densely spaced 802.11b transmitters will make it more difficult for receivers to distinguish desired signals from undesired ones. Hendricks fears that people will respond by trying to amplify (or otherwise boost) the 802.11b signal. Indeed, such hardware hacks already abound.

Hendricks points out that 802.11b equipment is certified to operate without a license, but only on a whole-system basis. Virtually every 802.11b hardware hack is illegal, he says. And this is only part of the destruction-by-popularity story. Other devices – like portable phones, Bluetooth devices, and (soon to come) radio-driven lighting – operate in the same 2.5GHz frequency band.

Hendricks thinks that 802.11b is a train-wreck in the making. Furthermore, he says, there is nothing to prevent 802.11a (also unlicensed, operating at 5+ GHz) from following a similar trajectory. As currently conceived, unlicensed spectrum could devolve into a hobbyist's playground. Independent network architect David P. Reed agrees, but he believes impending problems with unlicensed spectrum are tractable, given sufficiently advanced technology.

Reed believes that the key issue is scaleability. He points to Tim Shepard's 1995 Massachusetts Institute of Technology (MIT) thesis, now famous among high bit-rate connectivity fans, as support for Reed's insight that unlicensed data radios need only one additional property to become sufficiently scaleable to serve the general public – packet relay. Shepard's thesis demonstrated mathematically that a wireless packet-relay architecture could support tele-densities as thick as those found in Manhattan.

A packet-relay radio contains a radio and a router. Only some of the packets it receives have reached their intended destinations – the rest are forwarded to other packet-relay radios. Each packet-relay unit has some knowledge of its neighbors. Together, the aggregate of radios forms an ad hoc, self-organizing, multi-hop mesh network. In principle, service providers need only build access points within this mesh to, for example, connect to the Internet.

Wireless packet-relay networks solve the problem of multiple, powerful, overlapping transmitters. A network of weak transmitters (with routers attached) can send a packet a long way without unnecessarily trampling on the spectral commons. Multiple hops replace additional amplification.

Packet-relay radio networks have some other nice properties, too. They solve the line-of-sight problem that restricts single-hop 802.11b transmissions. Multiple hops can get around a large building or over a hill. In addition, packet relay does not have the problems of large, capital-intensive buildouts, because customers own most of the infrastructure. When you want to connect to a packet-relay network, you go down to Radios-R-Us, bring home a unit, and plug it in. When you connect, you beef up the network infrastructure – adding redundant routing and increasing the potential throughput of the entire network.

Economists call this "increasing returns." Reed calls it "architected cooperation."

Wireless packet-relay access remains an active sector, despite today's telecom recession. A few of the companies focusing on unlicensed spectrum include:

Will wireless packet-relay networks replace fiber access? Perhaps, but one can envision scenarios in which packet relay and fiber access grow together. For example, there may be a need for fiber in neighborhoods to support backhaul from packet-relay networks to Internet exchanges. Then packet-relay networks would form an infrastructure to support the discovery of bandwidth-hungry applications.

In later stages, the demand for bandwidth could grow so fast that an infrastructure of radios would be relatively expensive, compared to fiber access. At that point, communities (or entrepreneurs or forward-looking utilities) would build out fiber, because the need for cost-effective, high bit-rate connections would be stronger than ever.

— David S. Isenberg spent 12 years at Bell Labs before leaving in 1998 to found isen.com inc., a telecommunications analysis firm based in Westfield, N.J.