Full Service Network

~ the obituary ~

The Full Service Network went live on 14 december 1994 in suburban Orlando. Time Warner ran fiber into a neighborhood, hung coax off it for the last stretch, and put a terminal in each house that could pull video out of a server on request. Scientific-Atlanta built the boxes and Silicon Graphics built the servers. AT&T Network Systems, later Lucent, handled the network side, and the service area was aimed at 4,000 households.

It was not a mock-up. You could order a film and it started when you said so, with a working pause. You could order a pizza from Pizza Hut through the television, buy postage stamps, shop from catalogs, play games against other households, watch a soap episode you had missed, do your banking, print a full page in color through a Hewlett-Packard attachment, and read local news that responded to what you pressed. The guide that held it together was called GOtv. In 1994 this was closer to science fiction than to television.

The number that killed it is the terminal. Reports put the in-home unit at four and a half to five thousand dollars apiece, in 1994 money, per household. Time Warner Cable's chief technical officer Jim Chiddix has been plain about why since: the project was built to prove the thing was possible, not to prove it could be sold. It could not have been extended past its four thousand homes at any price a cable company could charge. Time Warner wound it down in 1997 and moved the lessons into a cheaper digital rollout.

December 1994, I was ten years old and four years away from my first modem. Nobody in the country I was in had video on demand, and neither did almost anybody in the country it was in, because there were four thousand of these and that was the entire population of the future. What I take from it now is a working reminder that most of the ideas we call new arrived early, ran correctly, and were shelved on cost alone. Streaming was not waiting on an invention. It was waiting for the hardware to get cheap.

~ the case file ~

~ what it could actually do ~

The technically serious piece was the video server. Silicon Graphics was building machines that could hold a library and serve individual streams to individual households at the same time, each one seekable, which is a very different problem from broadcasting one copy to everyone. Every architectural question that Netflix and YouTube would later answer at scale had to be answered here first at a scale of four thousand, on hardware where a single household's stream was expensive.

The commerce side is the part that reads strangest today, because it was right. Ordering food through a screen, buying goods from a catalog on the television, banking without going anywhere: all of that turned out to be what people wanted, and all of it eventually happened on a device Time Warner was not thinking about. The behaviors were correct and the screen was wrong.

~ the slide ~

Nothing about the trial failed technically, which is what makes it unusual in this graveyard. It was killed by arithmetic. A cable operator can spend perhaps a couple of hundred dollars putting equipment in a customer's home and hope to earn it back over some number of years. Five thousand dollars per home has no version of that calculation that works, and the components were not on a curve steep enough to fix it inside the window Time Warner had.

There was a second problem that had nothing to do with money. The whole system was custom. Custom terminals and custom servers, built for one deployment by companies working to a single contract. Everything that eventually delivered video to homes at scale was built from parts that already existed and were made by somebody else, which is the difference between a trial and an industry.

~ what replaced it ~

The direct descendant is boring and enormous. Time Warner took what it had learned and shipped digital cable with a set-top box that cost a couple of hundred dollars, and then cable video on demand, which became a normal product in millions of American homes years before broadband could have carried it. The Orlando engineers scattered into the companies that built that.

The Full Service Network gets written up periodically as a famous failure, which is a lazy reading of it. It answered the question it was built to answer. What it demonstrated was that the technology worked and the economics did not, and it is worth remembering that the people who ran it knew the difference at the time.

@}-->--

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

« previous  ·  random  ·  next »