Nova Scotia Credit Union League

We sold the Nova Scotia Union League a 9400 and a ‘banking' package called PROFITS (Personalized, Real-time Oriented Financial Institution Transaction System). PROFITS was cobbled together from a grab-bag of code written for some small American banks. It sounded good, but eventually we had to re-write most of it. The sale was made largely because of the personal relationship between the salesman and the DP manager of NSCUL.

The NSCUL PROFITS installation was an exciting project, my first telecommunications experience. The customer selected lower-priced NCR CRTs, so I had to write a driver that interfaced to the NCRs over a synchronous link. The 9400 had a huge diagnostic panel with hundreds of blinking neon lights, indicating the processor status. The communications controller also had lights, which proved crucial later in diagnosing a flaw in the communications protocol.

The disc drives, as big as washing machines, had glass covers through which you could see the movements of the arm that supported the read/write heads. This was useful because you could spot wasteful thrashing (rapid movement of the heads), which could be remedied by relocating the files to other drives to reduce contention.

At night, watching the silent behemoth lie in wait for data coming from afar, it was fascinating to watch the flicker of the communications controller signal the arrival of a transaction. The neon pixie lights would dance as the processor crunched the data, the disc arms would flail retrieving records from the ISAM master, another flicker of the comm. controller, and the data was gone. The CPU would resume its idle loop, waiting for another interrupt.

The computers were cooled by air flowing under the raised floor. We used rubber suction discs to lift the floor panels. If you knew how to spin the disc with a flick of your wrist, you could cause it to land flat and stick onto the glass walls that enclosed the computer centre. Waiting for a compilation to finish, we would while away the time by flinging these rubber discs from across the room, listening to the satisfying smuck as they clung to the wall. The next morning, we'd catch hell from the female operator who would notice the black rings on the glass panel.

In those days, the operating system OS/4 was shipped in source form. The first thing you did was customized and assembled the OS, called the "Supervisor". Reading the Supervisor's source listing was very instructive in learning the inner secrets of the machine. Unfortunately, some geniuses could not resist ‘improving' the Supervisor by modifying the code. Although their enhancements were often of value, when a new version of the Supervisor was shipped, the same hacks were not available (or not interested) to re-apply the patches, so the customer was stuck with an older version.

I heard a story at the Calgary office where an SA had added code to the 1100 Exec to count the number of times each magnetic tape was mounted. After a specified number of uses, a tape would be set aside for cleaning. This was very useful. However, because the code was difficult to port, the customer was relegated to running an obsolete version of the OS. Of course, Univac wouldn't respond to his bug reports because many had been fixed in later releases.

Calgary was at the heart petroleum industry, the "Oil Patch". The 1970's were the heyday of the Oil Boom. Univac sold enormous parallel processors, the size of a room, which did Fast Fourier Transforms and enabled geologists to analyze the seismic waves that revealed the location of oil deposits. Some of our clients were Texaco, Shell, Petro-Canada, ASGAS (Alberta & Southern Gas), AGTL (Alberta Gas Trunk Line), Occidental. I saw $100K (in 1970s dollars) decisions made over coffee break.

The Calgary branch had SAs from the Philippines, Honk Kong, and Sweden. When the Filipino SA became a father, he invited everyone to a party for the child's baptism. We roasted a pig on a spit. The Asians were gambling fanatics. We'd have these great parties where each dealer would choose the game. One customer, a kiwi, introduced us to ‘New Zealand whist'. I forget the rules, but we later figured out the odds were ridiculously slanted in the dealer's favour. It was almost like "If the card I deal you is red... OR black... I win!".

When Univac bought the RCA computer division, it was decided that RCA's more advanced VMOS (Virtual Memory Operating System), originally named TSOS (Time Sharing Operating System), would be used on the Series 90 computers. The converted OS would be called VS/9. Since both the Series 90, and RCA's Spectra series, were clones of the IBM 360/370 CPU architecture, it wasn't that much of a stretch. However, there were subtle differences, notably in the PSW (Program Status Word), which created a lot of confusion and difficulty. The PSW contains the result of the most recent comparison, the address of the next instruction to be executed, arithmetic overflow, and other vital bits.

The conversion of the OS involved switching the Univac hardware PSW to the RCA equivalent, and passing it on to the rest of the old VMOS code. It was often unclear whether the PSW was the raw Univac format, or the transformed RCA format, leading to much confusion and bugs.

I remember going on a VS/9 internals course. It was exciting, plumbing the mysterious depths of the OS. Much of the course consisted of training us on how to prepare dumps for analysis by the OS team at head office. Many crashes were caused by a module exceeding its allocated buffer and trampling on adjacent memory. There was a utility called SODA (Self-Organizing Dump Analyzer), to format the huge virtual memory dumps that were written to tape because they were too big to be printed. I set about writing a program called LEDA (Legible Executive Dump Analyzer), but did not complete it.

I believe one of the reasons that the Series 90 was not profitable is because of the huge expense of maintaining all the different OS teams (NCOS, OS/3, OS/4, VS/9, Exec 8, the 418 series, the 494, etc), and the unstability and complexity of an OS written in Assembler.

Univac also inherited from RCA a device called the CCM (Communications Control Module), a strange beast with a 17-bit word. The CCM was a large unit, which connected to Uniscope terminals. The CPU interfaced to the CCM through a continuous loop of I/O instructions, interrupted only when a terminal transmitted a block of screen data.

Although the Calgary staff were good guys, Calgary was primarily an 1100 town. I was transferred to Calgary to help promote the 90 Series, and I got a lot of ribbing about this upstart line of machines. There was constant rivalry between the 1100 supporters, and the Series 90 team. The two lines covered different ends of the spectrum, but there was enough of an overlap to cause confusion amongst the customers, and resentment amongst the staff who jockeyed for position, status and resources. Whenever a proposal was made to a customer, some SAs would be jealous if ‘their' machine was not offered.

This illustrates Univac's failure to educate its staff, and properly integrate the various lines of equipment.

Univac also manufactured the 1701 Verifying Punch, and the 1710 Verifying Interpreting Punch. The 1701/1710 featured an innovative blade feeding mechanism. 1701, incidentally, is the same model number as the Star Trek Enterprise.

As a learning exercise, I wrote a program that played the game of Battleship. The rules I had learned were somewhat more interesting: Each player calls out ten cells, but is only told which ships he hit, not where. Each time he hits an opponent's ship, he loses a shot.

The first version was written in Assembler on the 9200. The input device was the toggle switches on the panel. Results were displayed on the printer. This was in 1972.

When I was assigned to the PSBGM project, I converted the program to run on the 9400, and used the teletype console for input/output. Later, on a 90/60, I used a Uniscope terminal to interface with the player.

At Petro-Canada in Calgary (1978), I converted the program to PL/I on the 1106. I used HP terminals to play. I ran simulations all night to heuristically improve the program's strategies.

In 1994, I converted the program to C++, running under Windows.

My next project is to convert Battleship to Java, to run on the Internet.

Customer Engineers

Stories of maverick CEs abounded.

One time the machine had been down for several hours when the customer CEO walks in. He asks the Univac CE when the machine will be repaired. The CE looks at his watch and says: "When Mickey's hand is on the 12, it's a bench job".

Another time the DP manager, exasperated that the machine is down, says to the CE "Print something". In an amazing feat of hacking, the CE turns to the console, keys in a program in raw hexadecimal, presses a switch, and suddenly the printer starts spewing out paper covered with "Something Something Something Something Something ..."

At Maritime Life, a problem developed with our tape drives that the local CEs couldn't fix. An expert from the U.S. was dispatched. This guy with a long flowing cape and a Toulouse Lautrec hat shows up, doesn't say a word, starts working on the drives. At one point he brushed his gold watch against a 220 volt terminal and shorted it out. Finally, the tape drives are working. He leaves without saying a word. We boasted of having the only gold-plated Uniservos in the world.

Montreal

Around 1973, the Montreal branch of Univac had sold a 9400 to the Protestant School Board of Greater Montreal. The PSBGM wanted a time-sharing machine for its students to learn programming. The 9400 was a business machine, and did not have a time-sharing OS. Bob Milliken, a real genius, together with Tony Hutchins, adapted OS/4 to create UNIROS: Univac Real-time Operating System. They also converted a FORTRAN-clone from the Illinois Institute of Technology, IITRAN, to run interactively on the remote terminals. Unfortunately, they copied and sold the original IITRAN manuals as well. IIT sued Univac for copyright infringement, and won a significant judgement. This put an end to UNIROS.

I was assigned to do a massive conversion for Dylex, a Montreal textile manufacturer, together with another SA named Dave Sirken. Dave had a stomach ailment, and never ceased to remind us that he had a limited life span. We spent weeks on the night shift at Univac's Mississauga, Toronto Head Office. We became so alienated that, although we were staying at the same hotel, we each rented a separate vehicle to minimize contact.

The offices and computer rooms in Mississauga were guarded at night by a cranky old night watchman who insisted on locking all the doors. We couldn't get around to the washroom or the food machines because we were confined to the computer room. We would play tricks on him by wedging doors open.

One of VS/9's weaknesses was that the same terminals couldn't do applications and development: they had to be dedicated, and could only do one or the other. This did not impress prospective customers who were running DEC's operating systems.

Nova Scotia Light & Power

A 9300 sold to the Nova Scotia Light & Power was a small seed that led to a huge long-range commitment. NSLP (later NS Power Corp.) soon grew into a 9400, 90/70, then 1100s when Univac dropped support for the Series 90. At this point, NSLP could have switched to IBM because of the close similarity between IBM architecture, and the Series 90. However, Univac saved the account by donating massive amounts of manpower to convert to the 1100 technology.

NSLP collected utility payments with Optical Character Recognition cards. The OCR reader was a real-time device. Cards went through, were read, the computer had 35 milliseconds to verify the account number check-digit, and select the card to a reject pocket if there was a read error. A virtual memory OS was not suited for this type of application because of the delays caused by paging from disc. I wrote a real-time OCR program for VS/9. The program ran as a ‘Class 1', memory-resident program. This was exciting because I used the esoteric and privileged ‘SIO' (Start I/O) instruction, together with mysterious CCW (Channel Control Words) that directed data straight into memory, and interrupt handlers.

Once, glancing through the course catalogue, I decided that I wanted to visit San Francisco. There just happened to be an EXEC 8 JCL course at United Airlines. That's all the justification I needed to spend a week in a United aircraft hangar by day, and the exotic dancing establishments at night. San Francisco used to have some great belly-dancing joints.

Working at Univac was an educational experience. There were some bright people working there, as well as some eccentrics. It was a privilege to know them.

Strangely, I never received any sales or customer training. In contrast, when I worked for Xerox a few years later (on the high-speed 9700 laser printer), I spent three weeks in an El Segundo hotel room with twenty Systems Analysts and salesmen. We all spent one week on technical training, and two weeks on sales training.