A visit to Cambridge.

I haven't left yet. I've been reading about wireless networks.

It started with the cable problem. A telephone line would have to follow me through every passage and down every staircase. I'd be carrying reels of it, joining sections, trying not to pull a connection apart somewhere behind me. Ethernet cable would bring its own distance limits and the need for powered equipment along the route.

I was looking for another way when I found some of the work being done at MIT. That led me back to Rodney.

We used to work together. He's one of my closest friends, although I've let a lot of his letters go unanswered. His doctoral research concerned wireless networking, particularly how computers could remain connected without fixed cables. He teaches at MIT now and has kept developing the idea.

I remembered enough of our conversations to know what to look for. I read about Hyphos, Robert Poor's work at the Media Lab, then followed the references into routing protocols and experimental networks.^[1] I've spent days going through the papers, including the sections I initially skipped because I thought I understood them.

When I called Rodney, I told him I'd bought a large farm and was interested in sending information between different fields. Running cable across the property seemed impractical. I wondered whether he had anything I could look at.

He said I should come down.

That was most of the telephone conversation. I hadn't spoken to him in months, and I didn't want the first thing he heard from me to be a request for equipment.

I drove to Cambridge yesterday with my notes on the passenger seat.

---

Rodney met me at the Media Lab. We got coffee, and for a while we talked about people we used to work with. He asked how I was doing. I told him about the house and the repairs. He seemed relieved to hear I was occupied.

Then he showed me the radios.

They were less impressive to look at than I'd imagined. Circuit boards, battery connections, little antennas. Some of the enclosures had been opened and closed enough times that the screws no longer matched.

He had several units arranged around the room. He typed a message at one computer, and it appeared at another.

Each unit has a controller and its own power supply. It receives a message from a neighboring unit and passes it onward. The computer sending the message doesn't have to reach the destination directly. There just has to be a working sequence of shorter radio connections between them.

He disconnected one of the units.

The next message took longer.

Then it arrived.

He explained that another unit had been close enough to provide an alternative route. With a different arrangement, disconnecting that box would have broken the connection entirely.

We tried it again. This time nothing arrived, even after he reconnected it. He leaned over the table, pressed something on the board and muttered.

That was useful to see.

He has a working prototype, but some units lock up occasionally. The software sometimes keeps trying a connection that has stopped working. Recovering from a fault isn't always as simple as restoring power.

I took out my questions.

What sort of distance could I expect between units?

He wouldn't give me a single number. Outside, with clear space between antennas, one result. Inside a building, another. Placement mattered. So did whatever happened to be between the radios.

Would it work around a turn?

“What sort of turn?”

Around a building, I said.

Possibly.

What about several turns? Would I need a unit at each one?

He asked me to draw the buildings.

I drew a few rectangles and a line between them. He looked at it while I explained that there were several places on the property where I might want to collect data.

Then I asked about thick stone walls.

He put his coffee down.

“I thought this was between fields.”

There are old buildings too, I told him.

He wanted to know how thick the walls were. I said I hadn't measured them. He asked whether there was power in the buildings. I said I'd prefer to run everything from batteries.

He looked at the drawing again.

I wish I'd prepared that part better.

---

The papers had helped me understand the routing problem, at least.

In dynamic source routing, a request passes between neighboring units until it finds the destination, recording a route as it goes. The sender can then use that sequence of units to address its packets. If a connection fails, it can seek another route.^[2]

Rodney's demonstration made the limitation quite clear. An alternative route has to physically exist. If I leave a single chain of radios through the passages and one disappears from the middle, no amount of clever software will bridge a gap beyond their range.

I'd need to test a complete transmission back to the house before moving on. Record where I left each unit. Allow some overlap wherever possible.

Even with all the boxes present, messages can be lost. Radios interfere with one another, connections fluctuate, packets need to be sent again. I'd read the simulation papers, but watching him struggle with a few units in one room made their results feel rather different.^[3]

For a crude example, if each of ten links has a nine-in-ten chance of passing a packet on the first attempt, and those outcomes are independent, then the probability of the whole journey succeeding without a retry is

Acknowledgments and retries help. They also consume power.

That was where I brought up the timers.

I don't need a permanent connection. I could write on a portable computer, save the files and leave them waiting. Twice a day, every relay would wake up for thirty minutes. The network would establish its connections, pass whatever was waiting and send confirmation back.

Then everything would switch off again.

One hour of radio operation in every twenty-four.

In the simplest estimate,

If standby consumption is low enough, substantial battery packs could give me months of operation.

Rodney asked what I meant by “off.”

I said the radio would be off.

What about the controller? The timer? Would the unit keep its memory? Would it have to discover the whole network again every time it started?

We spent quite a while on that.

He agreed that scheduled operation could save a great deal of power. He wouldn't agree to months of battery life without measurements. He also pointed out that I'd have to account for the cold, battery capacity under load and the possibility of a unit spending an entire thirty-minute window trying to recover from a fault.

The timers would need to remain synchronized closely enough for neighboring units to be awake together. A relay waking just as the next one went to sleep would be useless.

Thirty minutes should leave some margin. Whether it leaves enough is something we'll have to test.

I've already had a watch stop down there. I kept that to myself.

---

I asked whether I could buy some of the units.

He said no. They weren't a product. The ones on the table were being used, and he couldn't simply let me take them home.

I asked about having another set assembled.

He wanted to know how many.

I said fifteen.

He laughed, then realized I was serious.

I told him I'd pay for the components and assembly. Whatever a separate batch would cost. I wasn't asking him to give me laboratory equipment.

He said I hadn't seen enough to know whether it would work on my property.

That was fair. I told him I understood the risk.

Then he asked why I couldn't just collect my measurements and bring them back to the house at the end of the day.

I couldn't come up with a good answer quickly enough.

I said some of the work would keep me away longer than that, and being able to send information back mattered to me. I told him it was important. I probably said that too many times.

He sat there for a while, turning one of the little boards over in his hands.

Eventually he said he would work out what a separate batch would require. About fifteen relays, experimental hardware, no promise that they'd behave properly outside the conditions we'd tested. The timer arrangement would need work too.

I wrote him a check before I left, toward the parts and assembly. He said he'd send me a breakdown once he'd priced everything.

I haven't brought any radios home. It'll take time.

---

The computer that actually posts these entries would stay in the basement, connected to the house telephone line.

I'd write on the portable and send a saved file through the relays. A program on the basement computer would receive it, make the Internet connection and upload the page.

Then it would send a confirmation back. Until I received that confirmation, the portable would keep its copy and try again during the next window. Each entry would need an identifying number so that a repeated transmission didn't publish the same page twice.

The tapes would still come home with me. This is for text. A few pages should be manageable if I can keep the connection working.

Fifteen relays may cover much less of the route than I hope. I won't know until I place them and test each link.

And all these calculations assume the equipment stays where I leave it.

I thought about the missing lamps on the drive home.

Rodney asked me to send him a proper drawing of the farm before the radios were ready. Distances, buildings, wall thicknesses. He wants to help me plan where they should go.

I still have the sketch we made at the lab. There's a coffee stain across one of the fields I invented.