Looking Up, from a Vision to a Spectrum
June 2, 2014
While I'm researching for another Newswire story, I've found an archive of reporting from the year that HP was taking its first full turn onto the path of RISC computing. RISC is the architecture that grew from the MPE XL version of the 3000 and its 900 Series systems, until finally HP evolved it into the Integrity lineup -- the only host that will ever run HP's Unix replacement OS. Back in 1985, it really looks like the company's CEO didn't know any more about 3000 designs than any other CEO at HP has since that time.
John Young was HP CEO, interviewed in the week while the Interex user group was hosting its Interex Washington DC conference. But the CEO wasn't at the conference. The company's founder was there, but David Packard wasn't the subject of the Computerworld interview. Young was asked what was prompting HP to pursue RISC as a computing strategy. He spent some time conflating and mixing several HP servers' technology. In the most baffling part of his answer, he said this about how muddled HP's computer architecture was -- and how RISC was going to change that.
We had desktops with one architecture, factory floor terminals with another and the HP 3000 with yet another stack architecure. The 9000 series terminals emulated the 3000 architecture in some ways, but not really completely.
Young went on to add that HP spent 90 percent of its development time changing things to make its networking perform correctly. "And those changes propagated down the whole computer line. I just decided, when I became HP president [in 1978]... that we wanted to find some way of bringing a harmony out of this unique business opportuntity. We needed to make a jump, and the conjunction of all those things was a program we Spectrum."
9000 series terminals? He probably meant the HP 9000 desktop systems, built for engineering. The 3000 architecture was Complex Instruction Set Computing (CISC), but so was the 9000's. Just a different design, called FOCUS. The factory floor terminals might have been attached to HP 1000s. One of the engineers on the scene at the time, Stan Sieler, told us he figures emulated in Young-speak might have been more philosophical than technological. Sieler also said that the sparkplug of RISC at HP was eager to get the Vision project out of the way, so Joel Birnbaum could enjoy his spectrum.
At that point in the era where the PC was only just starting to be a dominant business tool -- it now drives the largest share of HP's revenues -- and such computers were called micros, HP was sweeping technology away that it had spent years creating but never released. Failure was always HP's first option for the predecessors for Spectrum, Sieler said in his interpretation.
At one point, I was part of a task force that designed the "FOCUS-II,", which was pre-Vision (and pre-PA-RISC). It was supposed to be the next CPU architecture for the 3000, 9000, and 1000.
Scott Stallard was the chairman (he later became an Executive VP at HP), and others worked on it. But when we presented the report, we discovered that no one had told us we were supposed to fail -- so that Vision could be given the official blessing.
But neither FOCUS, FOCUS II, nor Vision were RISC CPUs. Birnbaum was hired away from IBM after Big Blue didn't want to create a RISC system, Birnbaum's dream design. Sieler went to work on Vision, then, only to learn that he'd been put on another blind alley. "I don't think that Vision fell short of what Spectrum became," he said.
To the contrary, it could do things that no subsequent architecture can. But, that came at a cost. Vision was definitely a CISC instruction set.
In 1983 (and somewhat earlier), I was doing design/development of process management for the HPE operating environment (for Vision). "Process management" meaning creating, starting, controlling, and killing processes (programs).
When I left HP (late September, 1983), we had one or two minimal (breadboarded) Vision computers running. Most of the time we used emulators/simulators involving re-microcoded HP 3000s. About a week after I left, HP killed Vision in favor of PA-RISC.
I once mentioned to Joel Birnbaum that it was cause/effect: I left HP, HP killed Vision. His response was quick: "If I'd known that, I'd have gotten rid of you earlier."
Sieler laughs at this today, bemused at the way things changed so quickly -- and then have not changed since. "HPE was renamed MPE XL," he said, "and most of the code written for it survived, To this day, much of process management in MPE/iX is still my code."