Stuart's comments on Metricom's spread-spectrum packet radio tranceivers

Note: I have no commercial connection with Metricom, other than as a user of their devices.

Wireless technologies seem to be falling into two broad camps:

The short range devices (a) typically run at 1-2 Mbit/s, and have a range of about 100 feet. Once you purchase the hardware and set it up in your office, there is no further usage charge -- just like when you set up an Ethernet or Token Ring network. These devices are often sold as "wireless ethernet", although they don't even go as fast as the original 3Mbit/sec ethernet (circa 1976), and they don't use the Ethernet CSMA/CD protocol either, so the justification for using that name is somewhat questionable.

The long range devices (b) typically run at the speed of a slow modem (2400 baud) and have the same coverage as cellular telephones or pagers -- ie. citywide, statewide, or countrywide, although there is frequently a high "roaming charge" if you stray outside your home area. In the future we can expect worldwide coverage via satellite services, though at an even higher cost. Like cellular telephones and pagers, there is a usage charge, either charged by the packet (a 230 character "message" in pager parlance) or by connect time (for the cellular telephone based technologies). Many of the wide area systems also have very long latencies -- for example CDPD can take many seconds or even minutes to deliver a packet, which is acceptable for e-mail but completely useless for a telnet login, for mounting a remote NFS file system, or for browsing the World Wide Web.

While there are lots of companies competing at both ends of this spectrum, Metricom seems to be alone in taking the middle ground:

Depending on your point of view, you can interpret this two ways:

I take the latter interpretation.

The "wireless ethernet" product class has little use for me -- if I'm inside the building and a few feet from a transmitter, I'd be just as happy to plug into a real ethernet connection. If you are going to equip your building with wireless transmitters everywhere, you could just as easily install multiple ethernet jacks in every room and provide a higher speed, more reliable service for less money.

The wide area cellular telephone and pager based services have little use for me simply because they are too expensive for me to afford, and if I could afford them they would still be too slow for the things I want to do.

With a Metricom radio, I have a connection that outperforms a 28.8kbit/sec modem (many home computer users still have to put up with less). I have sufficient range that I can go anywhere I like on Stanford Campus, in the surrounding area, and I've even heard reports of people using them as far away as San Francisco.

The radios can be operated in two modes:

I'm working on drivers for Linux to use the radios directly in the datagram mode. Performance results have improved by about a factor of three since our paper "Experiences with a Wireless Network in MosquitoNet" (see our Web Site). Of course it is is also possible to use the radios in the modem emulation mode using only the standard Linux SLIP driver, but the performance is as bad as a real modem. You would also need to provide a whole "modem pool" of radios for people to dial into instead of just a single Internet gateway, but if you are going to be the only wireless user on your system then I suppose that is workable.

I was recently at a talk organized by the Stanford Computer Forum, for Stanford Computer Science and Engineering faculty to present their current research work to local industry. I attended the two hour session where my advisor Mary Baker was giving her talk. I took my laptop computer and Metricom radio along, so that I could continue to get on with my work. I noticed that although there were other laptop computers visible in the room, I didn't see a single person using a cellular modem. Employees of many high-tech companies, including AT&T and Pacific Bell, were present, and their companies had shelled out $10,000 for the privilege of having them attend, so clearly shortage of cash was not a problem for these people. Why were none of these people using cellular modems? They are readily available over the counter in Fry's Electronics for under $300, so the problem is not that they are still research devices in the laboratory. Why don't I see people using them?

It also occurred to me that if there had been people with cellular modems in the audience, they wouldn't have used them the way I use my Metricom radio. When I looked up to listen to some interesting point, I didn't desperately hang up my connection because I couldn't afford the connect time charges, and then reconnect again afterwards. I left my Metricom radio "connected" for the whole duration of the session, and for the hour of discussion afterwards, and while I cycled home, and for the rest of the evening too. I'd like to meet the person who'll casually do that with their cellular telephone.

Of course the reason for this difference in usage is that the Metricom radios use a fundamentally better technology -- packet switching.

Even if cost were not an issue, you wouldn't leave your telephone "off-hook" all the time, because then no one else could call you. With a telephone, you are limited to one call at a time.

The reason that the cellular providers have to charge so much for air time, is because every second that you are connected is monopolizing limited resources in the network that other people need in order to make their calls, so there has to be some incentive for you to stop hogging those resources when you are not using them. With the Metricom radio, being "connected" doesn't consume any network resource or radio spectrum because it's just a virtual connection. Until you actually have some data to send or receive (e.g. pressing a key on the keyboard) you are not hogging any radio channel or preventing other people from using it, so there is no reason for Metricom to set up tariffs that encourage users to "hang-up" whenever possible.

To say that using Metricom radios has changed my life sounds too much like a religious conversion, but it has certainly made a big impact on the way I work.

You can find out more on the Metricom Home Page

The Linux Starmode Radio IP (STRIP) driver is available at

Page maintained by Stuart Cheshire
(Check out my latest construction project: Swimming pool by Swan Pools)