The Actual Size and Start of the 3000 World
September 12, 2012
Ask around to discover how large the 3000 world grew to be, or when it all began. You're likely to get a couple of stock answers. The size of the world counting every system ever sold might be reputed to be more than 50,000 servers shipped. Its genesis is regularly quoted at 1972, so this November would mark 40 years of the 3000's lifetime.
That genesis is only correct if you count the 3000s days of gestation. An out-of-print book by one of the longest-termed 3000 veterans, Thomas Harbron, tells a story of a often-patched, crashing computer. In Thinking Machines, Harbron reports that so pot-holed was the 3000's start that the delays gave HP enough time to rename the system the 3000CX, instead of the System/3000. As has been reported all over, the start was so flawed that HP's founder vowed no HP product would ever be announced before it could be demoed.
Harbron shared a section of his book with us that explains why HP needed the lengthy delay. (We'd love to bring Thinking Machines back into e-print, especially as part of the 3000 Memoir Project.)
After all, is it still a 3000 without a database or a business langauge? He says neither IMAGE or COBOL were ready to demonstrate until mid-1973. BASIC and the 3000's unique SPL were the only languages a customer could use to create software for most of 1973. (And creating your own software was the only way an application suite could go into work in the early 1970s on a 3000.) HP limited access to its lab 3000s in 1972 to communication via teletypes.
Harbron also reported that the first 3000 training seminar of two weeks in California was "classes that were mostly taught by Bob Green," the founder of Robelle. Even at the 3000's start, familiar faces of today were there spreading its seeds.
Harbron says in his book that things turned bad in a hurry when some of the other eight customers in that class tried to use that 3000 in December, 1972. This multi-user system couldn't support more than four simultaneous users. Time between crashes had a mean of about two hours, he says.
Every time I logged onto the system... a newer version of the operating system had been installed. In the two weeks I was there, the version number increased by more than 50, meaning there were that many versions in just two weeks. Clearly the developers were frantic. The effect on the people in the class was devastating.
While we were in the class, HP released a schedule for MPE. In it they acknowledged that they would not be able to deliver the full operating system on schedule. They showed an initial release in December with three subsequent releases which would bring MPE up to full specifications by June 1, 1973.
Harbron also notes that HP celebrated the shipment of the 25,000th HP 3000 in 1982. The company took a picture of the 1,000 people in the 3000 division and mailed this postcard to customers. Around that time the personal computer became such a serious alternative to business computing that HP began to sell the 3000 against it, even while it offered a touch-based console PC.
To continue those calculations, doubling that number of ever-shipped 3000s over the next 10 years -- before HP swung to Unix for preferred enterprise sales -- seems realistic. But the decade between 1992-2001 probably only showed give-and-take shipments. Losses of 3000s, some to HP's own efforts, which were then offset by new customers. Harbron published his book in 2000, showing the same kind of optimism we used in sizing up the 3000 market, as he looked back on that 1982 postcard.
The 3000 had become the fifth most popular computer ever made [by 1982]. Today, nearly 30 years after its advent, it is still selling briskly and has probably climbed even higher in the rankings (microcomputers excluded).