Monday, February 05, 2007

"Wireless Internet for All, Without the Towers"

Ok, Ok, I'm on it....

I just received my 4th and 5th emails pointing me to the recent NYTimes article "Wireless Internet for All, Without the Towers." Thanks! That's ample evidence that readers are interested in a topic I would have thought was off the beaten path here.

What motivates that interest is probably the prospect of a solution to the fabled "last mile" connectivity problem that relies on the "bottom up" efforts of individuals. That always attractive, at least ideologically, and when coupled with the promise of being radically cheaper to implement excites people.

Problem is, the writer's enthusiasm for individualism, dislike of municipal efforts, and perhaps a bit of naivete regarding points of physics and economics leads him to push a story about a system that can't be sustained and to miss the real story hidden in the hardware/software combination examined.

First, the naivete: The author, Stross, is touting a wireless mesh system. All such systems start from a connection to a network backbone that supplies bandwidth. Data is handed off between closely spaced access points whose coverage overlaps enough to provide redundant paths to downstream access points and to user's computers or other devices. The result is area-wide theory. Stross is right that this "area" system isn't working all that well, and that part of the problem is getting the signal inside the house where most use occurs can be spotty. Where he goes astray is in assuming that the network that he proposes can work even this well.

The Stross-proposed system implicitly starts from the interior of buildings with individuals or a small service provider making the intitial point of contact. The problem is that wifi signals have at least as hard a time getting out of house as getting in--and that relying on voluntary subscriptions and the resulting in randomly spaced nodes will make it irrational to expect that there will ever be enough nodes to mesh together effectively. Also, unhappily, it violates most bandwidth contracts to share bandwidth in this way. This really is not going to work as any sort of substitute for metro/area networks and implying that it might is misleading.

But there is good news. The story he misses is almost as interesting, and considerably more viable: As it currently stands this sort of system would be good alternative in apartment buildings. It could effectively deal with data and VOIP needs for apartments and dense public projects (but not video). In that situation a small ISP could provide legal bandwidth and control a stable, adequately dense, placement of access points at a minimized cost.

The best news is twofold: 1) The hardware is really cheap in comparison with other outdoor equipment. 2) Open source mesh networking software is maturing. The software that the Meraki system discussed runs is based on an older MIT project, RoofNet. CUWIN software, completely open source, runs on the Meraki hardware. Both projects have been supported by Google to at least the extent of providing software-writing interns.

Both developments promise to radically lower the cost for mesh networks in the long run. And lower costs will do more to make networking widely and usefully available than any fantasy.

Update 4:01: Jim Baller's always informative newsletter points to a muniwireless post that echoes some of the points made here, to wit:
The article, "Wireless Internet for All, Without the Towers," takes a roundabout road to discussing Meraki's offerings, talking first about the challenges of hanging mesh access points off lamp posts - relatively expensive, easily blocked by trees and buildings, etc. But Meraki's products today don't really fix those problems; you'll still need a street-based mesh, or some in-building Internet connection, to reach the wider Net. However, if you want to cheaply distribute a WiFi signal inside a building, this is a great solution....
It's nice to know that my contrary reservations are sometimes shared by the big dogs.

No comments: