| Early days | Maritime Life | Sperry Univac |
| NS Credit Union League | Xerox, COS | Scotiatech |
Tired of having to beg, borrow or steal computer time at Univac, I decided the only way I was going to do some serious system development was working for a company that had a computer and used it for running its business.
I applied for the job of Data Processing manager at Western Canada Hardware, a wholesaler and distributor of hardware (hammers, nails, drills, guns, etc). They didn't actually own a computer. They used a service bureau called Management Horizon Data Systems (MHDS), located somewhere in the Midwest. Data was transmitted through a data entry, Remote Job Entry (RJE) computer called a Data 100 or something.
The work was interesting. It included order entry, inventory control. We were restricted to what the MHDS system could do, which had its limitations. For example, every year we held a gun show where western retailers would place orders. The orders weren't to be delivered till months later, but we needed a sales tally that week. So we had to input all the orders, compute the totals, then promptly delete them, only to re-submit them months later from a backup tape. Nuts.
I toyed with the idea of developing a system on a Data General Eclipse system, but gave up when I realized WCH wouldn't support the costs. They were so cheap that the president disallowed air conditioning in the office. I once put a thermometer in the disc drive and measured 120°F. I got tired of arbitrating squabbles between data entry operators, and quit after about eight months.
Xerox Montreal was looking for a bilingual Systems Analyst for their new 9700 laser printing system. They paid my family's moving costs, and I moved to Beaconsfield.
I worked at Place du Canada, on the penthouse floor. Xerox was a strange outfit. They pioneered a lot of technology at the legendary Palo Alto Research Center (PARC), such as the mouse, LAN, and GUI workstations (the Alto), but never figured out how to make money from it.
I saw first hand how copier-centric Xerox was. Here was this exciting new technology, the laser printer. Yet the only thing the local management would focus on was copier sales. They assigned two dufuses to sell laser printers because they had "engineering" qualifications, from Vietnam and Morocco respectively.
These guys were so crafty that they managed to meet their copier sales quota, while maintaining lucrative sidelines. One guy had a restaurant where he employed his immigrant family, and also taught at and ran his karate studio. The other fellow owned a chocolate store specializing in expensive, imported Godiva.
This may have worked for small potatoes like copiers, but the huge 9700 laser printer was a $500,000 item (1979 dollars), and required a high-level, prolonged sales effort. The 9700 could print 120 impressions (60 pages) per minute, two-sided! The 9700 could churn out 50,000 pages (a whole pallet) per shift. It consisted of a PDP 11/34 computer running RSTS, a 50 MB disc drive, a 6250 BPI tape drive for input, and the print engine, including a laser the size of a coffin (an appropriate analogy considering how often the Customer Engineers (CE) had to replace it).
The 9700 was so fast it had dual input trays and dual output bins, so it could operate continuously. It would automatically flip to the second one as the operator serviced the first.
The salesmen were clearly not up to the task. When the Moroccan returned from a three-week vacation to the old country, he headed straight for his chocolate store, in spite of a thick stack of telephone messages awaiting him from anxious customers and prospects.
I grew very frustrated. My pride and self-image was tied to our sales. Xerox was so institutionalized that, in spite of the increasingly obvious incompetence of the laser printer salesmen, Xerox would not replace them until the end of the year, so rigid was their planning process.
We finally did get some competent professionals, Bill Herlihy and Guy Breault. However, the sales cycle was so long, and I had gotten antsy at almost a year of inactivity. At one point I was writing programs on a remote terminal to help salesmen compute optimal copier configurations for their sales proposals. I also wrote a program to play Mastermind (guess the colour of four hidden pegs). In addition, major corporations were reluctant to be the first customer, so we still hadn't made our first sale. When Howard April offered me a partnership, I jumped at an opportunity to form a company that would acquire a 9700 and offer laser printing services to Montreal companies.
Our company was called COS (Computer Output Services), later COS Information. Pierre Trencaroff joined us as Sales Manager, and both Pierres were minority shareholders. The next few years were difficult, as we struggled to develop a demand for this new service, and meet the $13,000/month machine rental.
Pierre
Trencaroff was a wonderful person, a real gentleman. He was a consummate salesman,
in the best sense of the word. Pierre always knew how to present an invoice
for unexpected expenses in the most positive light. The customer couldn't refuse.
His parents emigrated from Bulgaria, where they had experienced depressed economic
conditions after the war. His wife chided him for always keeping a 5 lb barrel
of sugar in his basement. He had known privation, when sugar was a scarce and
sought commodity. Pierre was a cultured person. He and his wife Suzanne, a real
sweetheart, loved opera. He had an extensive collection of classical music CDs.
He was fond of his old tube monoraul amplifier, and in spite of his technological
sophistication, refused to upgrade to stereo.
We slowly expanded, grabbing every opportunity however crazy it was. One of our customers was Sandy Sanderson, a tough, sleazy, underworld-type guy who boasted of connections with Premier Parizeau. His thing was stud books, showing the pedigrees of expensive standardbred horses. These books showed the race times of the dams and sires, and were used at auctions to evaluate the colts. They sold for $100+. We were offered the job, as long as we could print the unique ⅛, ⅜ etc characters used in measuring track lengths. Well, Xerox did not have a font that contained these characters, but they did have a digitizing service in El Segundo, CA, that could do customized characters, signatures, small drawings, etc. However, they needed artwork and a months's lead time. A month! Here it was Friday afternoon, and we had a customer who wanted thousands of books by Monday! Xerox had no appreciation for the small business' predicament.
This situation was the incentive to later develop our own font software, called COSMOS (COS MOSaic, after the bit maps that formed printed characters).
The 9700's rivals were the Honeywell Page Printing System (PPS), and the IBM 3800. The PPS was a pathetic dinosaur that used specially coated paper. The paper came in rolls, and was only manufactured at one plant in the World. I don't think Honeywell sold any of them.
The IBM 3800 became a very popular machine, but it was technically inferior to the 9700. It used fan-fold paper, and printed forms electronically through transparent glass masks that were custom-made for each report, and mounted by the operator before running a print job. The glass masks had to be manufactured by IBM, causing long lead times. The 3800 could only print on one side of the page.
In contrast, the 9700 used cut sheets, which were more versatile and less expensive. It used a programmable graphics language to draw lines around the report data. It had two input trays, allowing the use of coloured separators, or heavy cover stock. We were limited only by our imagination. Its appeal lay in reducing a big 11" × 15" computer report, down to a manageable 8½" × 11", two-sided. Salesmen used to joke that we "gave the customer half his desk back".
 9700
Sample output with electronic form - click to enlarge |
We were once approached by a mail-order company that had ordered 750,000 sheets of fan-fold intended for the 3800, only to find out that the stock was too heavy and wouldn't feed through the IBM printer. Desperate, they wanted to know if we could handle the job. We were equally desparate, so we accepted it. First, we had to burst or separate the pages, which were preprinted with yellow art work; this was done by a machine that pinched the paper and pulled it apart, which left a burred edge. Then we force-fed the paper through the 9700's trays.
The paper was so bad that we had to direct the exhaust from a vacuum cleaner into the stack of paper, to constantly flutter it and encourage the 9700 to accept the paper. After a couple of weeks, the burred edges of the paper had cut a groove in the feed rollers. The 9700 was filled with yellow dust. The Xerox technician said he had never seen anything like it.
The majority of our customers were regular companies and government organizations: Canadian Pacific, Gaz Métropolitain, Standard Life Assurance, the City of Montreal, McGill University, Alcan Aluminium. We did direct mail pieces for the Liberal Party, the Conservatives and the Parti Québecois, as well as General Motors. We printed order books featuring scanned bar codes for Provigo and Métro Richelieu, two big grocery chains.
We did a 200 copies of a 2,000-page concordance of the poet Virgil, the lifetime achievement of an Italin professor. The output was bound into a book using the perfect binding process. The paper is clamped into a carriage, dragged across a rotating metal brush which applies hot resin glues to the spine, and left to cool for a minute. The resin penetrates the gap between the edges of the sheets, and forms a strong bond. Paperback books are bound this way.
I sold during the day, and programmed and operated the printer at night. Paper jams required opening the print engine unit, and ripping out crumpled paper in eight different places. We used to have contest to see who could clear a jam the fastest. The photoreceptor drum was a razor-sharp metal belt, 12" wide and 10 ft. in circumference. It required cleaning every shift with a toxic corrosive liquid. More often if blemishes appeared on the printed paper, caused by anything from glue oozing out of pressure-sensitive adhesive stock, to a deteriorating laser. We had a catalog with samples of all the different flaws and their causes.
We knew that some customers were using us as a proving ground to test the technology, and determine if it would be suitable for their own, in-house use. They were up-front about it, and we accepted this, as some business was better than none.
The process of printing adhesive shelf labels was complex. Howard owned a printing company named Ampersand Data Graphics. Its specialty was offset printing on plastic sheets, mainly shelf price labels. After much research and experimentation, we developed the ability to laser-print bar codes on 'gummy label' stock. This was difficult because most adhesive papers would ooze glue under the heat and pressure of the xerographic process. The drops of glue would stick to the photoreceptor belt, and create black spots on all subsequent pages.
After printing a sheet of bar code labels, they were die-cut on a 1920 Heidelberg printing press at Ampersand, in an adjacent building. Sometimes the press would feed the sheet crooked, and the die-cutter would ruin it. The spoiled sheet numbers were written down, an operator would run back with the list, and frantically re-configure the 9700 to print the missing pages, called "remakes". This was done at night, (sometimes repeatedly), and had to be ready for the next morning.
One of our customers, Métro Richelieu, was convinced they would be better off if they printed in house, and decide to buy a 9700. We tried to convince them it was more difficult than they knew. In one of his last meetings with the customer, Pierre Trencaroff says he turned to the manager and said:
-"What about the remakes?"
So the guy answers:
-"What remakes?"
For the past several years, Ampersand used the services of a company called Logidec to typeset the master copy of the shelf labels that were Ampersand's specialty. Logidec had developed expertise in photo-typesetting, and unique software ("LOGIciel DE Composition"). They received tapes of data from our customers, such as Steinberg's or the Société des Alcools, processed them at McGill University, courier the output tape to their offices, where the job was typeset on an Autologic APS Micro 5. The large sheets were then delivered to Ampersand's office, where an operator made a photographic plate from the master, loaded the plate in the offset printing press, and printed onto plastic sheets.
Obviously Ampersand was strategically dependant on Logidec, since they were the only company that provided this specialized service. Imagine our panic when we discovered that Logidec had bought a 9700, and was planning on entering the laser printing business! (Ampersand provided the cash flow to fund money-losing COS).
We quickly decided that we would have to become self-sufficient: buy our own photo-typesetter, learn how to program it, and convert all our jobs. I went to New York in April 1983, at the annual Typesetting and Printing Trade Show. I checked out the options, decided the Autologic APS Micro 5 would work for us as well, and placed an order. Considering the $80,000 price tag, it was not a light decision.
Autologic typesetters were workhorses. Every major newspaper owned a couple of them.
When the Micro 5 came in, we wanted to preserve absolute secrecy so Logidec would not get wind of our plans. The same Xerox CEs serviced our equipment and theirs, and they might have talked. We destroyed all the shipping containers with tell-tale labels. We housed the APS Micro 5 in another building while we developed our typesetting software. We worked feverishly. Thanks to the outstanding work of our programmers, Marton Boros and José Rebatta, we were typesetting within a few weeks, using an Apollo Domain system as our computer.
Pierre Trencaroff described how he and Howard went to the Logidec to announce that we no longer needed their services. As they walked into the president's office, the president nervously asked if this was about the price increase. It turns out they had just recently sent us a notice to advise that they were increasing their prices. We had not received the notice yet, and were unaware. Pierre calmly said, "No, we're here to terminate our relationship." The look on the president's face was priceless.
In the years preceeding 1983, the advent of the Personal Computer, there was a proliferation of unique, special-purpose word processing machines: the Xerox 860, Micom, AES, Lanier, Wang, etc. Their business model was blown away by the PC and Wordstar, but in their day they played a significant role in office automation.
The word processors stored data on diskettes, each in their own proprietary format. Word processing documents were an obvious source of business, so we bought a specialized computer from a company in Tustin, CA (so cleverly did we disguise the name that I can't remember it), which could read the diskettes from most word processors. Wanting to keep this valuable discovery a secret from our competitor, we erased or taped over all the manufacturer's identification, and gave our machine the fictitious name "Wordtron" to throw spies off the track.
On an open-house trip to McGill, I discovered that the Electrical Engineering Dept. had an Optronics drum Photoscanner. I developed a process whereby we would make a negative of a customer's signature or logo, and send it to McGill for scanning. The scanned image would be copied to tape, and sent to the computer centre where we would convert it to Xerox' proprietary image format using software we had developed. A courier would bring it to our office in Lachine. We would then incorporate the digital file to the print job. We were the only company in the world with this capability, and sold the service to other companies.
I worked long hours, chasing the elusive dream of entrepreneurial wealth. It was hard on my family.
I had been experimenting
with decyphering the format of the font files. The 9700 included hundreds of
fonts, which were distributed on 1600 BPI magnetic tape. The fonts were strictly
bit maps, a long way from the True Type fonts used by Windows. Each combination
of typeface and size was a different file. The 9700 operating system included
a dump utility to print out the content of tapes, so I started dumping font
files and examining the hexadecimal data to look for patterns.
After pouring over many dumps, I began to see that the files were divided into consistent sections, with distinct patterns characterizing each. There were what looked like tables at the beginning, containing offsets to blocks further into the file. Using a patching utility, I poked random garbage in a copy of the file, and studied how it altered the shape of the letters. Sort of like nuclear physicists bombarding materials with particles and seeing what flies off.
I was making progress, and identified many elements, but was missing crucial information about the size or framing of the characters. One night, while poking around the Xerox cabinet which the CEs had left unlocked, I stumbled on a document that described the technical specifications of the 9700 print engine. There was an obscure reference to a 9/7 bit split of a 16-bit word. Eureka! That was it! The 9-bit part was the pixel width of the character (2^9 = 512), and 7 bits was the height (2^7 = 128). I was able to complete the model, much to my excitement and delight. We opened an account at the McGill University Computing Centre, and I wrote a program in Assembler to transform and customize 9700 font files. The program was called COSMOS, and sold for $5,000. Using special batch commands, COSMOS would set pixels to white or black, stretch or shrink characters, grey or stripe them, etc.
We sold COSMOS all over the world, including South Africa and France. We sold a copy to the French National Defense Department, universities, banks, Grumann Aerospace. One of our first customers was the Business division of Xerox in Rochester, New York. One of the conditions of sale was that we were not allowed to use them as a reference. The Rochester manager later told me that when he told Liz Bond, the Laser Printing Division manager, that he was going to buy COSMOS. She said "Oh noooo...". I later shared a dais with Ms. Bond at a Xerox User Group conference, and took great satisfaction in telling my fellow Xerox users that they didn't need Xerox anymore to develop custom fonts.
Compugraphic was a company that made phototypesetters. We visited Compugraphic in Massachusetts to try and interest them in licensing their extensive font library to Xerox users. Our COSMOS software would provide the translation. They were intimidated by Xerox, concerned about the legal implications, and declined. Compugraphic was bought by Agfa in 1989.
With typical secrecy, Xerox refused to disclose the format of the font files, even though we were paying hundreds of thousands of dollars per year in rental for the equipment. They also kept secret the metacodes that allowed fine-resolution control of character placement. I'll never forgive those bastards their obtuseness. This is why today, when you mention "laser printers", you think of HP, Lexmark, Epson, Brother, and others before Xerox. This ultimately backfired because if they had agreed to share the information in exchange for a non-disclosure agreement, we would not have been allowed to sell COSMOS.
It was particularly gratifying the day we sold special fonts to the Montreal branch of Xerox.
After five years, the business was still struggling. Howard April demonstrated bad faith by attempting to renegotiate our shareholder arrangement. The terms of our original agreement specified that Pierre Trencaroff and I had right of first refusal to purchase Howard's shares. This precluded the possibility of Howard selling his shares to an outside party, after we had struggled to grow the company. Well, clever Howard got a shyster lawyer to draft an amendment whereby Howard would sell his shares to his holding company, Lirpaco (April spelled backwards, get it?). He would thus observe the letter (but not the spirit) of our agreement by making Lirpaco the placeholder, and be free to sell Lirpaco shares to whomever he pleased.
We perfected the process of printing adhesive bar code labels, and were able to charge a premium to Métro Richelieu. The shelf labels were printed at COS, then processed at Ampersand as described above. The stock cost about 30¢/page, the laser printing went for about 5¢/page, and we sold the job for $1/page, a fair profit considering our development costs. Instead of giving COS the bulk of the profits, which was reasonable considering the enormous capital outlay, Howard rigged it so Ampersand paid COS 5¢/page, and pocketed the $1/page from Métro.
This duplicity, plus my divorce, prompted me to leave COS and return to Nova Scotia so I could be close to my children.
COS Information grew to 50 employees, and was sold to Vestcom.
DigiBarn Computer museum has more pictures about the 9700.
| << NS Credit Union League | Back to the Rants | Scotiatech >> |