Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ It's the Bandwidth, Stupid ]   |  Reach Out and Touch Someone  |   [ And Your Little Dog, Too ] >>

Weekly Column

Reach Out and Touch Someone: How Bob and His Binoculars Found More Bandwidth and Learned to Stop Worrying and Love the Bond

Status: [CLOSED]
By Robert X. Cringely
bob@cringely.com

My house sits on a hill in rural Sonoma County, 36,000 feet from the telephone company central office and 22,300 miles from the GE satellite that provides my Starband Internet service. I have written two previous columns about Starband, but this isn't a third one. Rather, it is a column about my feeble attempts to move beyond Starband and gain better service. Starband is nice, but the latency is real, the upload speeds are a joke, and some protocols can't be carried at all. Internet telephony, for example, works great, but only in one direction. So while I am not ready to dump Starband, I'm certainly looking for alternatives. And now I think I've found one — an 802.11b wireless link from Hell. Remember, I am a professional. Don't try this at home.

The whole idea was to piggyback off some other person's DSL connection. There had to be someone I could find who was close enough to the phone company central office for dependable DSL service, yet still within line-of-site from me. So I bought binoculars, then a telescope, then a larger telescope. In the early morning and late at night, I would sit on my deck scanning for neighbors with DSL potential. That's when the light was best and the haze was least. My immediate neighbors would be no help because they were all just as far from the phone company as I was. Most DSL needs to be no more than 18,000 feet from the C.O., so anyone else in Bennett Valley was off-limits. That left me peering through two little gaps on the shoulders of the mountain between my house and Santa Rosa. Through one gap I could see what looked like street lights, and through the other I could see what I was sure was a traffic signal, both between five and 10 miles away. To a man with slow bandwidth, such things are exciting. Enter telescope number three, finally big enough to actually see what was there — evidence of habitation!

It sounded easy. I'd find a house through one of those two gaps, knock on their door, then ask if I could buy them DSL service in exchange for mooching some bandwidth over a bootleg wireless link. It SOUNDED easy, but wasn't. People don't like to hear that others have found them by squinting through a telescope. They are suspicious of free offers. Or maybe they were just suspicious of me. But finally I had my co-conspirator, a guy who never would have bought DSL, but was certainly willing to enjoy it for free. We ordered Pacific Bell DSL because the phone company sometimes forgets to throttle the bandwidth, something a Covad or Rhythms would never do.

Now for the wireless link. I had done my research on the Net and knew exactly whom to copy, in this case a guy from New Zealand. The connection would be using 802.11b, which normally has a range of 100 feet or so. But that range is using an omnidirectional antenna. Using a 21 dB Yagi directional antenna, other people had built links as long as 14 kilometers while mine would be about 10. Interestingly, this sort of hacking doesn't appear to break any laws because the Effective Radiated Power of the Yagi is still under the maximum set by the FCC.

The set-up is simple. I had to buy a new Apple Airport hub for each end of the link. So I'm already $600 into it. Then I needed a pair of Apple G-4 internal antenna cables that matched the custom Airport coax connector. These cables are $30 each. Finally I needed a pair of $249 Yagi antennas and various cables, clamps, and poles, for a total of just under $1,400. Then came a little mounting, a lot of aiming, and the darned thing actually worked! Now for $49.95 per month I have close to a megabit of PacBell DSL to go with my Starband, all arbitrated by a multi-homed Linux Router Project homebuilt BGP-4 router built from an old 486/66 PC. At the other end of the link, no computer needs to be on for me to connect, just the Airport and the DSL modem.

What all this effort gets me is the ability to run a server. It gets me two-way Internet telephony, though at a cost where it would probably have been cheaper for me just to use my regular phone. It gets me AOL and several streaming protocols that just don't work very well over the satellite. But mainly it gets me two things I really value — reduced latency and a useable Internet time signal. I'll never have to set my computer clock again.

This use of 802.11b or WiFi is not all that unusual and probably indicates one direction where we will see commercial products headed soon. What I cobbled together for $1,400 could probably be put together in volume for half that price. I wish someone had done it and saved me the trouble.

This experiment is also a testament to the versatility of 802.11 — a protocol that was never intended to do anything even remotely like this. But 802.11 isn't what it used to be. There were a number of technical problems to be overcome along the way.

For one thing, the 802.11 protocol had some significant problems scaling up from LAN to WAN. First, there is the problem that, as the standard is written, every node on the network is supposed to be able to hear every other node so they can detect collisions much as CDMA Ethernet does. This "hidden node" problem — the node that is active on the network but can't be seen by the other clients — can lead to significantly reduced throughput if hidden nodes account for more than 10 percent of the total hosts.

Then there is the polling problem. 802.11 uses standard 802.3 polling, which means that a host first listens for traffic on the line, then transmits if none is heard. If a collision is detected, both colliding nodes wait a random time interval then try again. This works fine for Ethernet, but not for 802.11 because the signal strength can vary so much from one radio to another depending on where it is in the building, etc. So the 802.11 standard is very conservative with the result that it can take a lot of time for each computer to finally get a chance to talk. This is fine when the average LAN size is a few nodes, but when there are hundreds of nodes it's horrible — or was.

Finally, there is the packet size problem. Most of the 802.11 overhead goes into creating, then shutting down communication events. The actual data transfer is trivial. This suggests that the bigger the packets the more efficient the communication. Alas, 802.11 ramps up packet size in such a way that short packets are much more the rule than longer packets. And the spec even mandates that the first packets be as small as possible. This means another hit on throughput.

All of these problems were solved by Doug Karl, a guy who until recently helped to run the network at Ohio State University where all these wireless tricks were perfected. His company, Karlnet Communications solves these problems with its Turbocell technology, which uses the hub almost as a wireless switch, controlling for hidden nodes, maintaining polling priorities, and aggregating smaller packets into larger ones. Turbocell makes 802.11 far more useful than it really deserved to be, which is exactly the way Internet standards should evolve. Apple, Lucent, and other companies license Turbocell code for their products. So if you are reading this over a wireless connection at Starbuck's, be grateful to Doug Karl. I know I am.

But now what do I do with three telescopes?

Comments from the Tribe

Status: [CLOSED] read all comments (0)