Three Rivers Computer Corporation

~ the last words ~

Three Rivers Computer Corporation was founded in Pittsburgh in May 1974 by a group out of the engineering laboratory of Carnegie Mellon's computer science department. All of them had day jobs at the university and ran the company on the side. The first products were the lab's own projects turned into things you could order: converters built for speech and music research, and a vector display that could draw more than 50,000 vectors at a 60 Hz refresh rate.

The machine that made the name was the PERQ. Brian Rosen, one of the founders, had left in 1976 for Xerox PARC to work on a successor to the Alto, and came back in 1978 to start building a commercial version of the idea. The first order came in June 1979 from the Rutherford Appleton Laboratory in Britain, and the machine was launched that August at SIGGRAPH in Chicago. It is the first personal workstation with a graphical user interface that anyone actually sold. By the summer of 1980 the company had dropped everything else it made.

From 1981 the PERQ was developed jointly with ICL in Britain, and the company renamed itself PERQ Systems Corporation in 1984. Sun and Apollo arrived in the meantime with cheaper machines built on standard microprocessors. Rosen's own account of what went wrong is blunt about his company: it never had stable financing, and it waited far too long to move off its own hand-built processor. In September 1985 the investors declined to put in any more and PERQ Systems closed.

This one died the year after I was born, in a market made of research labs on another continent. The first machines I got my hands on were late 1990s PCs, running a windowed interface the PERQ had put in front of paying customers fifteen years before that. The Alto gets the credit for inventing the thing, and the company that got it onto a loading dock first does not get taught at all.

~ the record ~

~ what the perq actually was ~

The first PERQ delivered to Carnegie Mellon had 256 KB of memory and a 14 inch disk holding 27 MB. The display was the part that mattered: 768 by 1024 in black and white, taller than it was wide, at a time when a terminal meant 24 lines of green text. The processor was microcoded with a writable control store, so the instruction set itself could be rewritten in software, and it ran at a cycle period of 170 nanoseconds. Rosen's summary of the competitive landscape at that moment: Sun and Apollo were still dreams.

Its native instruction set was a dialect of Pascal's p-code called Q-Code, and the operating system was Accent, which came out of the same Carnegie Mellon research effort that later produced Mach. Rosen's own list of company mistakes includes following the university's operating system work too closely, which is how a commercial machine ended up running a research project.

~ the british second life ~

Bob Hopgood at the Rutherford Appleton Laboratory liked the PERQ enough to push the Science and Engineering Research Council toward adopting it for funded research. A British public body could not very well standardize on an American box, so SERC steered ICL into a joint effort to build PERQs in the United Kingdom, and ICL put money into Three Rivers along the way. The ICL PERQ ran in British university departments until the middle of the decade, under a common software base that SERC audited in 1985, the year the American parent ran out of road.

~ five dollars apiece ~

When PERQ Systems went under in September 1985, work on the unreleased PERQ-3 stopped where it stood. Carnegie Mellon had about 150 PERQs and no vendor, and there had already been a revolt among people forced to use them for their projects. Within a year most of those machines were gone from campus. A few went to former customers. The majority were sold to a local surplus dealer for five dollars each.

The remains scattered in several directions. Varityper bought the manufacturing rights to the PERQ T-2 and stopped making it shortly after. Crosfield Electronics in England bought the engineering designs for the unfinished PERQ-3 and funded two small companies to complete it, one of them Rosen's. A maintenance outfit staffed by former employees kept the installed base alive through a chain of mergers and renames until it went bankrupt and the landlord seized everything PERQ-related, and by 1991 the sign in that office window read "Woe Is Me, Inc." The company itself outlived all of it on paper: Three Rivers Computer was still listed in the Thomas Register in 1990, five years dead.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »