Pacific Connection(英語)

Community Wi-Fi Challenges 3G for High-Speed Wireless Access

The San Francisco Bay Area is one of the most "wired" places on earth. But when it comes to wireless data connectivity, Tim Pozar thinks the region has a ways to go. Pozar, a radio engineer, has worked with some friends to install some $2000 worth of 802.11b transceivers on nearby ridges. By commercial standards, this is a hobbyist project---but one with wide implications. The group has kept costs down through use of non-carrier class equipment, and networking the transceivers through point-to-point connections to a single broadband Internet feed. Pozar thinks this approach could eventually provide free or low-cost, high-speed wireless access to anyone in the area with a laptop or palmtop equipped with 802.11b ("Wi-Fi") access.

"Our ultimate goal is to create a network that isn't controlled by another company, to which you may have to pay a recurring fee," says Pozar. "With wireless, after you pay for the up-front costs, you have no recurring fees except for rental and maintenance costs. At the scale of a neighborhood, you just put up an external antenna and an Internet access point and all your neighbors can plug into your network. They could use this for sharing files back and forth or sharing resources, like printers. Or if you went out and bought a non-encumbered broadband connection, you share that with your neighbors, as well."

Pozar's vision is one of the most ambitious versions of what some are calling a metropolitan area network, or MAN---whose proponents argue is the most promising way to get a high-speed wireless connection at the lowest cost. "Look at 3G: they're going to be spending billions of dollars for equipment and spectrum," he says. "By contrast, with 802.11, you could become your own little ISP that's located in your little corner of the city. You're not going to be offering the same feature sets---for instance, with 3G you'll be able to roam anywhere. But 3G's bandwidth is much smaller. They say 2MB, but nobody is going to get that. I think it will be closer to 100 Kbps.

To the north in Washington State, Seattle Wireless, another not-for-profit community project, describes MANs as community-owned, distributed wireless access systems. MANs are "NYASPTWYOMB - not yet another service provider to whom you owe monthly bills." That attribute separates them from wireless voice, which is metered by the minute, as well as from commercial purveyors of 802.11b access.

An under-hyped protocol

Today, no true metropolitan area network exists, and the groups proposing them admit the idea is still highly experimental. But what's already clear is that the 802.11b protocol was a rare case of a technology being under-estimated and under-hyped by its inventors. Officially released by the IEEE in June 1997 by the same group that defined 802.3 Ethernet, 802.11 was envisioned strictly as a wireless version of a local area network. The practical uses included transmitting information to employees like hospital nurses, who were mobile within their workplace. The 147-company Wireless Ethernet Compatibility Alliance consortium launched in August 1999 with the mandate of ensuring compatibility. The group coined the term "Wi-Fi"---for wireless fidelity. The Wi-Fi logo acts as a kind of "seal of approval."

Not surprisingly, the most frequent Wi-Fi deployments are for home networks, as represented by Apple's AirPort product. Encased in white plastic and resembling a child's top, the AirPort base station can function within a 150 foot radius and support up to 50 users, making it perfect for a classroom setting. Similarly, The Intel Wireless Gateway supports up to 32 PCs, 16 wired, 16 wireless, which Intel markets as part of its AnyPoint Network for home and small office. Altogether, some 75 vendors make equipment branded with the Wi-Fi seals, including Sony, Fujitsu, IBM, Compaq, Cisco, Epson, LG, Philips, Sharp and Yamaha. The attraction is clear: one Internet connection, whether dialup, cable or DSL, with no messy cables to run along corridors, under rugs, or beneath the floor.

But while 802.11b wireless LANs are supposedly restricted to about 200 feet, there's no law of physics that says it must stay within the user's walls. Signal "leakage" is inevitable and that has lead to inadvertent expansion. Take David Sarno, who, according to the New York Times, moved into his apartment, ordered a high-speed Internet connection, and prepared for several weeks delay. Then he discovered that his laptop, equipped with an 802.11 card, could "borrow" a neighbors connection. It's debatable whether his connection was fully legal, although in truth, the network owner could secure the connection against casual use through WEP (Wired Equivalent Privacy) encryption, among other techniques. But as a short-term parasitic act, his seemed innocent enough. Nicholas Negroponte, the director of the M.I.T Media Laboratory, has cited this kind of bandwidth exchange as a useful byproduct of 802.11b connectivity. You use my network connection when you're near my office, I use yours when I'm in your neighborhood. Such a network connection is potentially similar to other infrastructure service, including roads and streetlights. As taxpayers, we provide these services to locals and visitors alike in the expectation that we will be served in kind when we take to the road.

Of course, it was inevitable that people wouldn't leave such connectivity to happenstance. On the website of is a map of the U.S. that plots network connectivity for the country ( . California, especially San Francisco and Los Angeles, are well covered, as is the Eastern Seaboard. There are islands of connectivity in Houston and Florida, and chains of connectivity stretching along some of the Interstate highways. There is even some connectivity Havana. Much of this research has been conducted by so-called "war drivers," who can link their laptop to a GPS, load the appropriate software, and drive off in search of unsecured connections. NetStumbler itself is one such program and its users are both security professionals, and hobbyists who want to map connections. A comparable program for BSD is dstumbler (

If wireless networks can create inadvertent, if involuntary, connectivity, others can do so explicitly, either charging users for the privilege of a high-speed wireless connection or using such a connection to attract customers. These so-called "hot spots" include airports, coffee shops and Internet cafes. In a report by HereUare Communications issued last March, a company that helps companies build their own hot spots, Starbucks coffee was the wireless access champion, with 539 locations. Cafes and restaurants accounted for 55 percent of all public Wi-Fi services, hotels accounted for another 38 percent. It's a good bet that wireless connectivity will be an expected part of airport facilities. As cell phone ownership becomes all but universal, Wi-Fi services will become more important than pay telephones.

A faster "last mile"

While the convenience of wireless access has been widely touted, one of 802.11's biggest advantages is its high bandwidth. Even the slowest of the 802.11 versions, 802.11b, can theoretically support speeds of up to 11Mbps (although actual speeds are about half that), while 802.11a and 802.11g versions theoretically support speeds up to 54Mbps. As a result, 802.11 is a promising medium speeding up the "last mile" connection---the perennial bottleneck of Internet connectivity.

That Brewster Kahle's motivation for creating a network of antennas around his house in the Presidio area of San Francisco, a deactivated army base that is now overseen by the National Park Service. (George Lucas is planning to move several of his businesses to the area.) Kahle, who founded the Internet Archives, has arranged to put up about six .5 watt transceivers to cover a square mile area---one mounted at the nearby San Francisco Exploratorium, another at a fire station, a few on to people's roofs. So far, the service is intermittent. "At this point, it's strictly a proof of concept," Kahle says.

For Kahle, the "concept" is to get Internet pricing tracking onto the Moore's Law curve. Semiconductor performance is supposed to double every 18 months. "But there are no plans to double today's bandwidth every 18 months, delivering twice the performance for the $40 a month Americans are now paying for high-bandwidth services," Kahle says. And why are faster speeds needed? Kahle would like to see bandwidths capable of movies on the Internet: not the choppy, mini-screen streaming video of today, but full-screen, satellite-quality video. Kahle calls current DSL speeds of 500-700 Kbps "painful." What's needed is about 1 to 1.5 Mbps for newscast quality transmission, 3-5 Mbps "to transmit Demi Moore at qualities comparable to a satellite service," which in the U.S. provides the highest quality film distribution outside of DVD.

Kahle thinks that the problem goes beyond ISPs trying to find new services and revenue streams---affecting the rest of the computer industry. For example, hard drive capacities continue to surge, but there's no motivation to buy them, because people have nothing that large to store. Film storage could be that motivation. And then there's the fascinating uncertainty about this new channel of film distribution. "The Internet created a wild explosion in the publishing industry," he said. "I'd like to see the same thing happen to moving images."

Kahle's 802.11 project might be termed a neighborhood area network (NAN), a small cousin of Pozar's MAN. But even at this small scale, Kahle is not fully satisfied with the results. "It has worked pretty well, but not great," he says. "The problem is that the equipment being sold is designed for the wireless LAN market. You can't blame them for that, but this is still consumer-grade stuff."

The other problem is spectrum and power." Kahle says that the unlicensed spectrum used by 802.11 is a good thing, a required ingredient in the generation of new ideas. But he says that the Federal Communications Commission, which oversees spectrum allocation in the U.S., did it for the wrong reasons. "The FCC is in business of selling spectrum, and left this bit of spectrum property to the free world because nobody would buy it. It's the worst." As builders of 802.11b networks often note, the 2.4 Ghz frequency is the same one used by microwave ovens, which heat water by oscillating the molecules. The result: a lot of noise. "It's as if all of the U.S. got sold, except for a nasty part of Nevada [the least populated state, known for its high deserts]. It's on that little sliver of land that we're going to build the new world."

The cost of free networks

Beyond the neighborhood area network is are broader networking projects, called variously "MANs," "community networks," or "free networks." A list published on's website shows projects in various states from early conception to in-progress, throughout the U.S. and Europe, as well as in Australia, Argentina and Peru. (For reasons I'm unclear on, no comparable Asian projects are listed.) These projects have a nice historical resonance dating back to the now-legendary Homebrew Computer Club, whose participants went on to found 23 companies. In piecing together a community network, the hardware hacker/cracker community is again working on a potentially "disruptive" technology---one that fundamentally alters the workaday world. Like open source on the software side, community networks provide an alternative to commercial offerings that is lower in cost and controlled by the people who use it.

The network conceived by Tim Pozar is an outgrowth of his association with the Bay Area Wireless Users Group (BAWUG), which is devoted to education. Pozar's friend, Matt Peterson, founded BAWUG as an educational endeavor after creating a high-speed network for the Burning Man event in Nevada. Pozar's MAN would use the ridges above the Bay Area to provide 802.11 coverage throughout the region---from San Jose in the south to Solano county in the north, and beyond---providing speeds of roughly 5Mbps.

But how can you do that with technology designed to connect a house or office, where the typical network base station doesn't promise to transmit beyond 150-200 feet? "The FCC doesn't restrict on distance, they restrict on how much power you are actually putting out," Pozar said. The FCC rules depend on the type of equipment, whether it is point-to-point---one transmitter, one receiver, using directional equipment---or point to multipoint using omni directional equipment.

For wireless LANs, broadcasting beyond 200 feet is more of a security disadvantage, and is hardly needed to illuminate a household. Consequently, base station manufacturers use just a fraction of the 1 watt allowed. "If you actually try to transmit at the maximum levels that the FCC gives you on a point to multi-point arrangement, you can actually do about eight kilometers. If you do a point-to-point arrangement, you could probably do 20 to 30 miles," Pozar says. His scheme would do both: linking the antennas through a point-to-point connection and reaching users via point-to-multipoint. The Bay Area's mountains make this approach viable, and Pozar and friends are still experimenting to see how well the 802.11b protocol works in the Bay Area's terrain. The first link uses two transceivers to span the San Francisco Bay, running between the city of Hayward, south of Berkeley, to Sign Hill in the industrial town of South San Francisco.

A different approach is being used in areas of the country that lack surrounding hills. In Illinois, the Champaign-Urbana Grassroots Wireless Internet Project wants to erect a series of rooftop, omni directional transceivers, each of which would serve a small cluster of houses. Each transceiver in turn would be linked to the others through a directional antenna, constructed out of a Pringles potato chip can. (The design has been posted and critiqued on the Web).

"Our biggest goal is to keep the price-point as small as possible," says Zachary C. Miller, a systems engineer for the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign. "We want to set members of the community with a router and a standard software distribution." The router would make the decision about which house to send a given packet. As envisioned, the system would have some redundancy, so that if one antenna was down, data would still flow.

While broadband Internet connectivity would be offered, Miller sees home-grown community advantages as well, including real-time streaming of community video programs, alternative e-mail services, and voice-over-IP. Other useful applications may not be apparent until the network is built. "Everything about telecommunications had been controlled by corporations," Miller says. In a high-tech throwback to the pioneers who farmed the prairie, wireless connectivity represents a new form of an old tradition: community self-reliance.

While self-reliance does not necessarily mean "free," it does mean that high-bandwidth could be distributed for low costs. Miller contemplates perhaps a couple of hundred dollars per household, although the project is still too new to determine the actual cost. Pozar says that the cheapest part is the equipment: transceivers are a one-time cost and can be hand-built. "We're using spectrum that is effectively free at this point, and the gear is so commoditized that deploying an access point would be about $500 or even cheaper." The real cost is in ongoing charges. Depending on the number of users and their speed expectations, the bandwidth demands at the Internet access point could be considerable. Today, the charges are between $200 and $500 per Mbps per month.

Another ongoing cost, potentially larger, is the rental fees for the antenna sites. Pozar has managed to get some mountaintop locations donated. "But if I had to pay for that, it could be between $500 and $2,000 a month per site. Mountaintops cost money. Bandwidth costs money. You really need a sugar daddy to make it free. I'm not ignoring these facts." Pozar says that while he and his cohorts are paying for the experimental network themselves, he's not averse to a government grant. He'd also like to see community wireless network design become more systematized, so that others could build them without having to start at square one. For example, Pozar does wireless installations in Indian reservations in North Dakota, and sees 802.11 as a low-cost alternative to the $40,000 to $100,000 now spent for carrier-class equipment at each site. "Native Americans are living below the poverty line so they don't have a lot of cash to spend on this. But they definitely need the infrastructure," he says.

Still, the term "free network" shouldn't be taken too literally. "To me, it means making the pipe to the Internet as cheap as possible," Pozar says. "We may also be able to make the pipe between users of the network free, depending on how much people try to donate to the cause. If we could get all of our mountaintops donated for free, if we could get all the replacement equipment donated for free, if we could get all the bandwidth we needed donated for free, then the service will be too cheap to meter."