Background
After my Antarctic adventure (see Weather Radar on a PDP-8 in Antarctica), I arrived back in Boston in July, 1978. Hanging out at the MIT AI Lab as I had done before Antarctica, I ran into my friend Radia Perlman, who asked what was next for me. I replied that I was looking for a job for the summer, and would go back to the 'Tute in the fall. She said “Oh, you should go work with Mike Speciner at Camex!” I said, “What’s that?”. [See below.]
They looked at my resume, which had (I think) two relevant bits. One was that I was someone who could be shipped off to Antarctica with stone knives and bearskins and succeed anyway, a flexibility important to a small startup company. The other was that, in lieu of a degree I didn’t have, I mentioned that my undergraduate advisor was Ed Fredkin, not knowing that he was the founder of III, another pioneering computer graphics company. That was seemingly enough; I was hired. There were six employees at that time, so I got to wear just about every hat: shipping and receiving, assembly and field service, in addition to my nominal title of “System Designer” which actually meant “junior software developer”, at least to start. I ended up working there for nine years, and never earned a degree, though I still hung out with friends at MIT.
Camex
Camex was founded in 1974 by Tom Hagan, an MIT engineer, and George White, a Harvard grad. Their vision was to produce electronic systems for newspapers to transition from the manual cut-and-paste method of preparing ad copy for publication to what was known as “electronic pre-press”. Using a computer to design and print a complex page with different type fonts and graphics is unremarkable today, but it was totally unknown in 1974 save for the very first Xerox Alto machines, the existence of which were basically known only to Xerox PARC employees. They saw a market in helping big-city newspapers compose display ads using a computer workstation rather than paper, scissors and glue.
Tom believed in “the big picture”: everyone, including junior new hires like me, should be aware of the whole company and its relationships with customers, suppliers, etc. I adopted that worldview too, and practiced it at all the companies I worked at and positions I worked in.
The Camex 135
The first product was the Camex 135. (George White had worked at IBM, where all product names were numerical, and he brought that attitude with him. The first customer, the Boston Globe, was then located at 135 Morrissey Blvd.) Since the goal was to handle a full broadsheet newspaper page, 17" x 22", it needed a big display and a big tablet for users to draw graphics on. The display CRT was about 25” in diameter and very deep. The tablet was also the primary command interface: an array of “buttons” were printed on a piece of paper about 8" x 10" which was taped to the tablet. The tablet was about 36" x 30", plenty of room for a full-size newspaper page and the command table. A full keyboard was tucked under the monitor, and could be taken out if text entry or editing was required. The front end (display processor, tablet, keyboard) and display smarts came from a PDP-11/05 running a dedicated application. The application maintained an in-memory model of the ad, which was accessed via DMA by the vector display engine made by Adage, Tom Hagan’s previous company. The vector display had a bit-slice microcoded processor of its own, to read the display list and drive the analog electronics that controlled the CRT. It had a built-in character display ability, but only one “font” and it was unable to display variable-width characters. However, the application software knew about fonts and displaced characters to give a good idea of the actual font height and character widths.
The backend was driven by a PDP-11/35 running CAMEXEC. DEC’s operating systems were considered inadequate, so the company rolled its own. System software (CAMEXEC itself, DDT, TECO, the 11/05 application, the text composition software and the APS-4 driver) was stored on a 5Mb cartridge disk, compatible with the DEC RX-05 but made by a cheaper third party. The 11/35 and 11/05 communicated via DDCMP over a parallel interface. Text was typed on dumb terminals and transferred in on 8" RX01 floppy disks, which were also the way ads were saved. There was no floppy file system, each ad used one floppy disk. The output was sent over RS-232 to an Autologic APS-4 typesetter.
The text composition software was translated from PDP-6/10 code written by Lowell Hawkinson for Composition Technology Inc., a commercial computer typesetter of mathematics Mike Speciner had worked for in 1973-4. Dick McQuillan, who was president of CTI, did the original translation.
Aside: it turns out to be remarkably hard to draw vectors on a CRT screen with constant brightness. See The Secret Life of XY Monitors for a discussion worthy of its own rathole. Adage had solved this problem as well, though I’m unable to find any description of their solution. What I do remember was that the analog circuitry had to be tuned from time to time as components aged, via a half-dozen potentiometers at the edge of one of the circuit boards. Naively turning one would distort the effects of the others, so I learned to adjust them in a set order and iterate as required until the display was acceptable.
Aside: Mike Speciner also included in his driver software the ability to do graphics, by decomposing them scanline by scanline and drawing rules (typography-speak for “lines”) for each scanline. This made the typesetter run abysmally slow, and I suspect it wasn’t used much in production for that reason. Later posts will talk about how later products handled graphics efficiently.
Boston Globe
By the time I joined, there were three Camex 135s: the development and demo unit in our office, and two in production at the Boston Globe. The Globe had several filing cabinets where they would keep the floppies for each ad, paperclipped to a copy of the typesetter output. When the same advertiser ran a similar ad the next day or week, the version on the floppy could be restored and modified.
New York Daily News
Our next sale was six units to the New York Daily News. They were the same as the Globe units, except they used the then-new PDP11/34 computers, made expressly for OEMs like Camex. I played a large part in assembling, shipping and installing those units, as well as writing the typesetter driver for the Linotron 505 typesetters at the Daily News.
The Daily News at that time was a hybrid operation. The middle of the paper, with most of the ads and non-time-sensitive matter, was phototypeset, while the front pages with late-breaking news were still set using hot-metal Linotype machines. It was fascinating to see the pages composed in this manner, which were very heavy, wheeled to the pressroom on carts that were built of heavy steel and must have weighed several hundred pounds each.
That was also my first lesson in disruptive technologies. The Daily News was a union shop. Linotype machines were just that: machines, with motors, clutches, gears, cams, levers, etc., and it took trained machinists to operate and maintain them. So the machinist’s union, having been in charge of the Linotype machines, demanded that they be in charge of the replacement technology: the phototypesetters and electronic pre-press systems including Camex’s. The “job” was the same, even though the technology was utterly different and so the skills of the machinists were almost entirely irrelevant. We had to be careful to work with the union people, and I saw first hand how they were afraid that this new tech would make their livelihood obsolete.
Washington Star
The final customer for the 135 was the Washington Star, which purchased two units. I assembled, shipped and installed those units as well. The pressroom was adjacent to the room where I installed the 135s, and it was interesting to see the press operators communicate via sign language. Half of them were deaf, not a disability in a job where one couldn’t hear spoken words anyway, and the others learned out of necessity.
Another adventure at the Star happened the first day that I worked past 4:30. I was vaguely aware that the others working in that room had all left, but didn’t think anything of it until I heard an earsplitting noise. It turns out that room was also home to the 500-character-per-second papertape punch that received the daily stock ticker dump. It spooled hundreds of feet of papertape onto the floor in a heap about the size of a small beanbag chair. It made a sound like having your skull cut open with a chainsaw; no wonder everyone left the room! I quickly learned to do likewise. The papertape was then fed into the APS phototypesetter and discarded. That rigamarole was later replaced by a Xitron box which had an input card replacing the papertape punch, a memory card, and an RS-232 interface to the phototypesetter. No noise, no consumables!
Alas, the Star went bankrupt not long afterward.
Epilogue
I have no idea what happened to the 135s once they were phased out of production; I assume they were junked. With total production of eleven Camex 135s, of which ten were sold, it was clear the long-term survival of the company required a different approach. I’ll describe that in my next posting.
I’m indebted to Mike Speciner for reminding me of some important facts in this posting.
