[Previous entry: "Russ Nelson is a crank"] [Next entry: "Polls Show Many Americans are Simply Dumber Than Bush"]

02/08/2006: "ESR, Russ Nelson and 802.11 based Mesh Networks"

What ESR doesn't understand about RF *and* the 802.11 MAC speaks volumes. I'm sure Eric means well, and he and I agree (shock shock!) on the solution being more (open) competition with the ILECs.

Quoting:

Wireless mesh networking — flocks of cheap WiFi nodes that automatically discover neighboring nodes and act as routers — is the technology that can do that. With the right software, networks of these can be self-configuring and self-repairing. It’s pure libertarianism cast in silicon, a perfectly decentralist bottom-up solution that could replace wirelines and the politico-economic choke-point they imply.

The main thing holding wireless mesh networking back is the small size of the bandwidth now allotted to it for spread-spectrum frequency hopping. With enough volume, competition would drive the price of these creatures to $20 or less per unit — low enough for individuals and community organizations to spot them everywhere there’s an electrical grid. Increments of capacity would be cheap, too; with the right software, your WiFi card could aggregate the bandwidth for as many nodes as there happen to be in radio range.

Can scalable RF-based mesh networks be built? Yes, absolutely. Information theoretic proof is available.
Can scalable 802.11 based mesh networks be built? No, absolutely not. The 802.11 MAC won't scale. Information theoretic proof also available.

I gave a design outline for a scalable fenceline based network in my blog a couple months ago. It uses a 'fiber down the fenceline' approach with FSO (google for "Ronja") to cross the street. A few geeks could build such a network, and it could scale hundreds of city blocks using mostly off-the-shelf technology. The resulting network would have far more capacity than any imaginable wireless network, and cost far less to build.


Matt: increased range is the enemy of scalable mesh networks.

Russell also gets it wrong, and (sadly) by so doing, further spreads already oft-repeated misinformation.

the reason that open source wireless drivers are hard to get is because software these days is mostly defined by software. Thus, with the right driver, you could transmit on any frequency that your output stage can handle. The FCC gets remarkably grumpy about that and won’t give permission for completely open source wireless drivers. Same thing for winmodems. Of course, a libertarian will recognize this as the all too visible dead hand of government.


The FCC doesn't care if your HAL or driver is open or not. The FCC cares if a manufacturer of equipment sells (or offers to sell) equipment that isn't compliant with FCC regulations. (Go study at the commerce clause in the US Constitution if you don't understand the difference.) The chipset vendors restrict the publication of HAL and/or driver source code via contract. They do so in order to comply with FCC regulations, which state, in part:

§ 15.15 General technical requirements.
(a) An intentional or unintentional radiator shall be constructed in accordance with good engineering design and manufacturing practice. Emanations from the device shall be suppressed as much as practicable, but in no case shall the emanations exceed the levels specified in these rules.

(b) An intentional or unintentional radiator must be constructed such that the adjustments of any control that is readily accessible by or intended to be accessible to the user will not cause operation of the device in violation of the regulations.

Between 'a' and 'b', you're screwed (at the FCC layer) if you violate the out of band emission standards of Part 15 (using a device you hope to certify under part 15 rules.)

If you don't like it, its a free market, and you are free to scrape some money together and bulld (or buy an existing/abandoned) 802.11 MAC and PHY, then have UMC or TSMC (or perhaps IBM) build the chipset you've designed (or purchased) in volume with *whatever* software you choose.

Such might be an uneconomic decision for you, since few (if any) ODMs would use a chipset that couldn't comply with 'b' (above), and this would result in a huge loss for your company. If you're a publicly traded company (as are the major 802.11 chipset vendors), your stockholders will cry 'foul' when you spend on R&D with no hope of breaking even, much less making a profit.

Still, its a free market, and you're free to try. Quite literally there is nothing to stop you. Unless, of course, you want to import those devices from whence they were manufactured, or ship them interstate, or even offer them for sale out of state.

Remember, they're *chipset*, and its unlikely that there is a foundry in Malven, PA (Eric) or Potsdam, NY (Russ).

You could put a couple FPGAs down on a board and actually build a real software defined WIFI router, but it wouldn't have a COGS low enough for you to even begin to sell them for $20 (or even $2000) and make a profit.