Chapter 7 - Data Communications: Market Order 1973-1979
7.3 Codex: LSI modems and Front-End Processors 1973
In early 1973, Carr found himself in unfamiliar territory. Codex was profitable, demand for its modems never stronger, and the 1972 year-end public offering had provided sorely needed cash. Existing contracts and changes in lease accounting virtually guaranteed near-term revenue growth. Thus freed from the stresses of having to manage every day as if it might be their last, Carr turned his attention to the threat of LSI semiconductor technology and decided it was time to contact Rockwell, a firm he knew was interested in modems from their having tried to sell Codex the failing ADS.
Exploratory telephone conversations confirmed Rockwell’s interest and motivations. They wanted to fabricate a 9600 bps modem chip to sell to Japanese facsimile manufacturers.27 Rockwell also made it clear they had no intention of selling stand-alone modems. The opportunity seemed too good to be true, so Carr, Storey and Forney arranged to visit Rockwell. In a pre-meeting strategy session over breakfast at the Anaheim, California Disneyland Hotel, Forney told Carr and Storey his Rockwell contact made clear that they had no interest in a relationship with Codex unless Codex committed to buy a minimum of 5,000 chips. How easily hope turned to despair. The idea seemed so preposterous that they could hardly maintain eye contact: 5,000 chips meant 5,000 modems when only eighteen months earlier they all remembered laughing so hard they cried when Pugh projected selling 400 9600 bps modems. Fear, however, makes the impossible seem doable, especially for someone like Carr who had endured too much to fail now. Carr could not dismiss the rumors that Milgo, and others, had LSI modems under development. So rather than write the trip off as a valiant effort, Carr proposed:
Look, if we go in and we tell this guy that we’ll commit to 10,000 and he buys the deal, we’ll get started.” So we went in and we had the meeting and we committed to 10,000 modems, and he took it with a pretty straight face, and I gave it to him with a straight face, and we cut a deal.
After months of negotiating a contract, including the thorny issue of circumventing antitrust regulations, the difficult and time-consuming work of designing and fabricating the chip began.28 Carr could now do little but wait to see if he had averted disaster.
Carr’s immersion in the world of semiconductors gave him valuable insight into the rapidly advancing state of semiconductor technology and the availability of commercial chips. Discussing this knowledge with Pugh in one of their frequent end-of-the-day conversations when as friends they quit being boss and subordinate and sat with their feet perched on whoever’s desk, to speculate as to what their competition was up to or what the latest customer trends were or what they should be doing to survive and prosper in a fast moving industry in which no one had claim to a crystal ball. As usual, they wondered if it was time to start development of a computer front-end communication processor Their 3C experiences – when they sold minicomputers as communication front-ends for mainframe computers – informed their instincts. Might substituting microprocessors for minicomputers yield a new product line, a line that might sustain Codex if modems in fact proved doomed? Developing computer technology competencies would also force Codex to overcome a deficiency that concerned both men.
Then, as if a prayer had been answered, James Vander Mey, a professor at the University of Illinois, asked Jim Rothrock at a Codex trade show booth if they had any interest in front-end communications products. Rothrock forwarded the inquiry to Forney who, after confirming the legitimacy of the source, discussed the issue with Carr. Soon Carr and Forney were making trips to Illinois to sell Vander Mey on joining Codex. Vander Mey wanted to be acquired, but after the Holsinger experience, Carr strongly preferred hiring people and avoiding the baggage of an existing organization. Carr’s persuasiveness won the day, and in May 1973, Vander Mey and a handful of his best students became Codex employees.
In 1973, the state-of-the-art of front-end communication processors, while generally understood, remained in implementation proprietary to each manufacturer. This posed a challenge for Forney since Vander Mey reported to him. How could Forney master these new technologies that were deep in content and superficial in exposition? Reviewing the available technical literature pointed him to recognized experts such as Wesley W. Chu, Professor of Computer Science, University of California, Los Angeles (UCLA).29 Chu pioneered work in a new technique of communication line-sharing known as statistical multiplexing (SM), or asynchronous time-division multiplexing. While his work held only tangential importance to front-end processors, it directly affected Codex’s TDM’s which were based on more dated techniques of synchronous time-division multiplexing. Forney discovered more relevant literature, but as events would unfold, Chu’s work would have the most impact.
While Forney investigated the technology, Pugh began assembling a marketing team by raiding Memorex, the only company offering a competitive product to IBM’s best selling 2703 front-end processor. Soon marketing was specifying product requirements for engineering and writing a business plan to drive strategy and operations. Not long into the process, however, Pugh concluded he and Carr had significantly underestimated the commitment required to compete against IBM. For one thing, Codex had no experience creating and supporting host computer software. A cautious Pugh recommended all development be stopped and, before the end of the year, Carr concurred. By then, however, Vander Mey and his team had so impressed Forney, Pugh and Carr that they wanted to find another project for them to work on. They choice was a statistical multiplexer.
Collins Electronics, an acquisition, did not have the requisite 9600 bps expertise.
Each party wanted to prevent the other from selling the modem chips to its respective competitors, such as Rockwell selling the chips to Milgo.
The interested reader is referred to: Dixon R. Doll, “Multiplexing and Concentration,” Proceedings of the IEEE, vol. 60, pp. 1313-1321, Nov. 1972. Chu’s original paper: “Design considerations of statistical multiplexers,” Proc. 1st ACM Symp. Probl. Optimization Data Commun. Syst. (Pine Mountain, GA), Oct. 1969