Kell carries key account of 3000 revival
July 16, 2014
We've come to learn that community icon Jeff Kell is battling a serious illness. While I wish this keystone of MPE wisdom a quick recovery, and the best wishes to his wife, I'd like to share some insights he relayed about the transition from Classic 3000s to the ultimate edition of the server he's worked on and cared for most of his career at the University of Tennessee at Chattanooga.
I'd asked Kell to explain what the HP CEO during that transition era, John Young, might have been talking about while the CEO told Computerworld in 1985 about the strategy of RISC. As the clipping from Computerworld to the left shows, Young was a lot less than clear about what RISC would do for HP's long-term computing plans. A comment in the second paragraph of the clipping -- about networking, one of Kell's most ardent studies -- made me want to reach out to him earlier this summer. Young's conflation of "9000 series terminals emulated the 3000 architecture in some ways, but not really completely" was something Kell could clear up.
I'm not aware of any similarities [Young noted] between 3000/9000 Series except after adoption of RISC, and they used the same processors/hardware. They may have shared some peripheral hardware earlier, but certainly had little in common until RISC. The 3000/9000 had practically nothing in common prior to that other than perhaps HP-IB peripherals.
Network-wise, the 9000-series was following the ARPA/Ethernet track, while the 3000 initially started down the IEEE/OSI architecture. Ethernet was only accepted by the 3000 as an afterthought, it was a checkbox on the NMCONFIG dialogue if you wanted to allow it, and it defaulted to OFF.
So unless Young was talking post-RISC (timeframe is wrong), I'm not sure how he would compare 3000/9000 lines at all. The initial RISC 3000s were in the last half of the 1980s. If I recall correctly, my "migration training" to the "new" 3000s was at the Atlanta response center around 1985 (or a little later) and we were expecting a 930. We ended up with a 950 (since the 930 sucked so badly.) But I do recall many of the details.
"At that time," he said, "we had stretched our Series-IIIs to the limit. HP had "loaned" us a 42 and 48 to "tie us over" until delivery of Spectrum. We had the week at the migration center in Atlanta and spent most of it doing switch stubs for our extensive set of SPL support routines. We finally got a 950 and never looked back, but we had several engineers scratching their heads in the process. We were doing some really peculiar stuff."
Those were "interesting times" indeed. I think at the time we had a Series III with 64 terminals attached (production), a Series III-R (development), a Series 40 or 42 (Library), an academic 44/48, a leftover Series III (academic), and that loaner pair of 42/48 (or 52/58?) to tie us over until Spectrum. We were long overdue for an upgrade, but no hardware was available yet to satisfy the need.
The 3000's direction on networking was most disturbing, taking the OSI standard model in the midst of our evolving Sun/Solaris Internet computers. We had 3000s on our LAN that could only talk to other 3000s on our LAN... while the rest of the server room was on the Internet. It was laughable.
It would be another decade before Posix came to MPE, and it started to play well with the other kids on the block. But unfortunately, a decade too late.
HP executives were taken up with the "Unix" movement... and the 9000s dominated their focus. The 3000s were just along for the ride. And looking back today, that wasn't such a great bet either.
Kell is a classic example of a chapter of living history -- and the lessons we learn from it -- that should be cherished by the community. After nearly 40 years, the decommissioned 3000s at his UTC shop were picked up for recycling. "We're now officially 3000 history," he said, "with nothing left on site."