
Traf-O-Data
» see traf-o-data.com as it lived, on the wayback machine
~ the goodbye ~
In 1972 traffic engineers counted cars with a rubber hose laid across the road and a box on a pole. Every time an axle crossed the hose the box punched a 16-bit pattern into a roll of paper tape, recording the time and the count, and somebody then had to turn that roll into something a city could read. Firms in Seattle were doing the work by hand. Two people who thought they could do it cheaper were a Lakeside School student named Bill Gates and his friend Paul Allen, who had just left for Washington State.
They called it Traf-O-Data. The first version was manual labor: read the hole patterns off the tape by eye, then push the transcribed cards through a machine at the University of Washington to produce a traffic flow chart. Intel announced the 8008 that same year and Allen worked out that a small computer could read the tapes directly and skip the human step. Neither of them could build hardware, so they went looking around the University of Washington physics building and found an electrical engineering student named Paul Gilbert who could. He became the third partner and built the machine.
Allen scraped together $360 for the 8008 itself. The finished prototype landed near $1,500 two years on. Then a man from King County came to see it work, and it did not work. The partnership carried on processing tapes for a handful of customers, two counties near Seattle and one district in British Columbia, until the State of Washington began offering the same processing to cities for nothing. Receipts across the entire life of the business were in the thousands of dollars against a larger loss.
All of this was finished four years before I was born, on a continent I would not see for decades, and there is no version of it I get to claim. What I can recognize is one move inside it. Allen wrote a program on a PDP-10 that pretended to be an 8008, so the software could be written and tested before Gilbert's machine existed. Anyone who has had to build against hardware still sitting on somebody else's bench knows exactly what that is and why you do it, and it is the reason this story is worth anything at all.
~ the chronicle ~
| Born | 1972 in Seattle, out of the Lakeside Programmers Group |
|---|---|
| Peak | a working 8008 machine in 1974, built on roughly a $1,500 budget |
| Killed | 1980, by which point Washington State was processing traffic tapes for cities at no charge |
| Lifespan | eight years |
| Made by | Bill Gates and Paul Allen, with Paul Gilbert on the hardware |
| Killed by | a state agency giving away the service they were selling |
~ the box they built ~
Gilbert had a working microcomputer running on the Intel 8008 in 1972, which puts it among the earliest machines anyone assembled around a microprocessor for a paying purpose. His brother Miles, a draftsman, drew the company logo. The 8008 ran at 500 kHz and reached 16 kilobytes of memory, and Intel was selling it to people building terminals and calculators rather than computers, so treating it as the processor of an actual machine was an argument rather than an obvious step.
The demo failure is the part everyone repeats, and what makes it worth repeating is that the machine had worked before. It read tapes correctly in testing and then stopped in front of the customer, which is the oldest failure mode in the trade and no less humiliating for its age. Allen wrote later that Traf-O-Data was a good idea with a flawed business model, a generous reading of a demonstration that died on the table.
~ the funny part ~
The business that failed was the traffic business. The skill that came out of it was the one Microsoft got built on three years later. When MITS advertised the Altair 8800 in january 1975, Gates and Allen wrote a BASIC interpreter for a machine neither of them had ever touched and watched it run correctly the first time it was loaded in Albuquerque. They could risk that because they had already done the same trick for the 8008 and knew the method held.
So the accounting on Traf-O-Data reads as a failure while the technical ledger does not. A few thousand dollars of receipts against a net loss, and an emulator that turned into the first product of the most valuable software company of the twentieth century.
~ the inheritance ~
Nobody agrees on when the partnership actually stopped. The Wikipedia infobox puts it at 1975. Allen's own figures run from 1974 to 1980 and land on gross receipts of $6,631 against net losses of $3,494, which means somebody was still filing paperwork for Traf-O-Data at the end of the decade. The likeliest reading is that it stopped being a business years before it stopped being a name on a form.
Type traf-o-data.com into a browser today and you arrive at microsoft.com. The domain belongs to the company that the failure produced, and it does not bother to serve a page explaining why.
~ leave a rose ~
an anonymous tribute. one rose per visitor per day.
@}-->-- ... roses left here
~ leave a wreath ~
a rose fades by morning. a wreath stays on this grave for good, with your name on it.
from $1, paid once through stripe. it keeps the graveyard running: the hosting and the nightly digging. adopt it and the headstone reads "maintained in memory by" you. the groundskeeper plants new wreaths at the next tending, usually within a day.
» one email a week: new graves and the anniversaries of the dead. subscribe
other software: Project Looking Glass, Adobe FormsCentral, Java Applets, HyperTerminal
also died in the 1980s: Imagine Software, Betamovie, System Development Corporation
~ leave a note ~
visitors before you have left these graveside notes. anonymous welcome.