Chapter 14 - Internetworking: Emergence 1985-1988
In 1981, Proteon was one of the start-ups contributing to the Emergent phase of Networking. (See Chapter 8.4 Proteon) In June 1981, Proteon shipped its initial product: the ProNET-10, a 10 megabit token passing LAN. Since most experts expected IBM to introduce token-passing LANs as well, until they made their products clearly known, the market uncertainty hovering over token passing made life very difficult for Proteon. Finally, on October 15, 1985, IBM announced a 4M bps token passing LAN – not one of the alternative technologies, especially Ethernet. Nevertheless, during those intervening years the “Ethernet” companies had gained market advantage and scale. Understanding these dynamics are essential to understanding why and how Proteon entered the Internetworking Market in 1985.
Important as well to this story is the close relationship Proteon enjoyed with MIT’s Laboratory for Computer Science (LCS). Proteon developed the ProNET-10 under contract from the LCS and funded by DARPA. DARPA was keen to understand whether the deterministic performance of token ring was superior to that of the stochastic Ethernet. At the same time, DEC had gifted a large number of VAX computers to LCS to be networked together. MIT chose to use Proteon’s token ring LAN(s) as the backbones. Unfortunately for Proteon, DEC elected to sell Ethernet, not token passing, LANs to its commercial customers. That same year, Mike Dertouzos, director of LCS, and friend of Proteon’s founder Howard Salwen, introduced Salwen to Noel Chiappa, a research scientist with LCS. Chiappa becomes important in the history of multi-protocol gateways/routers. The bond between Proteon and LCS was further reinforced that year when Prof. David Clark of LCS – and TCP/IP fame - joined Proteon’s Board of Directors, undoubtedly with the blessings of Dertouzos.
Picking up the story in 1983, it was the period during which Proteon was struggling to generate the cash it needed through operations. While raising money from venture capitalists seemed possible, Salwen was unwilling to cede management control to those he viewed as more interested in “flipping” his company – preferring a quick return rather than building a “real” company - that was until he met Jon Bayless, a partner of the venture capital firm Sevin Rosen Bayless Borovoy. Bayless had begun calling on the company in early 1983. Even then Salwen politely shunned Bayless’s overtures; that was until a project undertaken in August threatened the company’s existence. Salwen remembers:
People were interested in our technology, but had no way to use it. We could do Multibus and Unibus and Q-bus, and that was it, but they liked the idea of the token passing ring. A fellow from National Advanced Systems, Larry Corey, insisted that we do a PC card and do it now. I had no money; I mean I was strangulating for lack of funds. You have to understand this, and yet we committed to him to do a PC card, which led to P-1300. We started in August and we delivered in November.
Then the serendipitous: Novell, the leading LAN operating system company, contacted Proteon. Salwen remembers:
They wanted their software to run on everything that ran, and when they tried our PC adapter card, that they were able to do in mid-November of ‘83, they were really impressed with the speed. They had tested nothing like it. I mean it just flew. And so, they gave us part of their booth at the November ‘83 Comdex tradeshow. Ray Noorda, President of Novell, helped us tremendously by bringing people over to the booth. I can remember John Doerr, of the preeminent venture capital firm, Kleiner Perkins Caufield & Byers, standing there talking to me, and Ray Noorda, again and again, with his arm around somebody, saying: “Now, I’d like to show you something really special.” And Doerr is watching what’s going on and had a real appreciation for what we were doing.
By December 1983, the stress of juggling insufficient cash between payroll and accounts payable, and never having enough to invest in market opportunities finally coming their way, Salwen accepted the inevitable. He agreed to terms and raised $2.4 million of Convertible Preferred Stock Series A. In doing so, he ceded management to the investors and agreed to increase the burn rate by hiring a management team. (Investors were led by two of the most coveted partnerships: Sevin Rosen Bayless Borovoy and Kleiner Perkins Caufield & Byers.) Salwen afterwards devoted himself to engineering, LAN standards activities and Board responsibilities.
A search for a new President and VP of Finance began immediately: a temporary President was installed. While the investors had confidence in token ring, they needed a strategy until IBM made its intentions known. Core to defining their strategy was agreeing to a product development plan. Proteon’s plan, of course, required making a best guess as to when IBM would make known their LAN strategy. The most valuable input to that decision was the pace IBM was forcing within IEEE 802 to finalize a token ring LAN standard. After considerable discussion, the best guess was 1985 at the earliest. Given that the token ring market was not going to take off before IBM’s endorsement, then either the market continued to lumber until 1986 or other technologies would be adopted by customers – a sobering outlook.
Three product development strategies emerged: continue to improve the ProNET-10 product line and prepare for IBM’s presumed token ring choice; introduce a high-speed token-ring backbone LAN, a product essential to their university and campus customers; and develop products that made interconnection with other LAN technologies more effective and efficient. During 1984, Proteon’s product announcements reflected primarily incremental innovations to the ProNET-10.
Also that year, the personal relationship between Salwen and Mike Dertouzos, the DIrector of MIT’s LCS bore unexpected fruit. Upon returning from sabbatical, Noel Chiappa a LCS research scientist, shared an insight for a multi-protocol gateway dreamed on warm sand beaches, with Dertouzos. Dertouzos authorized Chiappa to begin work on the code for what would become known as the “MIT C gateway” and introduced Chiappa to Salwen – for the charter of the LCS did not include building, much less marketing, products.81 Chiappa was soon working for both organizations. At MIT, before the end of the year, Chiappa’s vision led to Internetworking router software running in all of the DEC MicroVAX IIs linking Ethernet to the token-ring backbone.82
In March 1985, Proteon introduced their second product, their backbone LAN: the ProNET-80, “the first commercially available local area network to operate at speeds up to 80M bps.”83 With this product, Proteon distanced itself from the other would-be token ring competitors.
Finally, on October 15, 1985, IBM announced its Token-Ring Network.84 As expected, the Token-Ring Network conformed to the IEEE 802.5 LAN standards. What was slightly surprising was its maximum 4M bps speed, or data rate, given the 10M bps speed of existing Ethernet LANs and of competitor token ring LANs. IBM also left unclear how the earlier announced IBM PC Network (the Sytek product) and the Token-Ring Network would coexist. Everyone assumed the Sytek product would be quickly phased out, but IBM would not make its intentions known. While other vendors took weeks to respond, Proteon announced its ProNET-4 the same day as IBM made its announcement. The ProNET-4 came with not only the IBM PC interface but also Q-bus. Multibus, VMEbus and IBM-compatible microcomputer interfaces. No matter what one thought of the IBM announcement, all agreed a giant dark cloud had been lifted and there should be no more excuses for why the token passing LAN market did not take off. And it would!
On that same day, Proteon established early leadership in the emerging Internetworking market with the introduction of the p4200 LAN gateway. The p4200 could interconnect its new ProNET-4, and its ProNET-10 and -80 networks as well as Ethernet LANs and public WANs.85 The ProNET-Linkway software for the p4200 was derived from Chiappa’s vision and the working MIT C code. The p4200 is considered the first commercially available multi-protocol router. (The use of the term router came only over time; finally taking hold in 1987. The p4200 was known first as a LAN gateway then a Gateway router and finally Router) Proteon further broadened its product line with new passive ProNET Wire Centers that supported a wide range of media connections including shielded twisted pair, coaxial cable, and optical fiber. A diagram published in DataPro in September 1986 is an example of Proteon’s product offerings. (Exhibit 12.14 ProNet LAN Configurations)
Exhibit 14.18.1 ProNET LAN Configurations
In December 1985, the arduous search for a new President came to a close. Francis Sciricco, a young, fast track General Electric executive, then managing a consumer appliance division in Kentucky, made the leap to the still promising, but struggling technology company. The desperate need for a President had out weighted his lack of technology experience, and thus judgment. He faced the daunting tasks of bringing executive order to the systems and structure of a company no longer a start-up but too often acting as such, as well as the urgent need to refine the development and marketing strategies to fit markets and technologies in the throes of great change. The chaos within the “corner office” hardly had time to settle when a disastrous fourth quarter precipitated change throughout the management ranks, including the resignation of Sciricco in early 1987.
L.J. Sevin, lead investor and Board Member, assumed the thankless role of Acting President. No one knew better than he that the company’s future rested on the completion of a timely search, and this time, on successfully finding a fully qualified technology executive. Within days of settling into the flow of operational decision-making, the highly regarded Sevin concluded the same as had management: the Networking market was best described as tumultuous. It was if the competition had been secretly preparing during the years before the IBM announcement, and now that IBM had announced what could be disparagingly described as unimpressive products, competition unleashed a barrage if products to gain market share before IBM had time to exercise its market power. Customers too seemed more willing to get on with building out their networks, no longer fearing that IBM might either introduce a radically superior technology or release undeniable proof that LANs were some great fraud. The Ethernet companies, with the compelling advantage of multiple chip suppliers and strong balance sheets from recent IPOs (Ungermann-Bass, 3Com, Bridge Communications) or mergers into larger competitors (3Com merger with Bridge Communications, Tandem Computers acquisition of UB, Interlan into Micom or Fox Associates into Digital Communications Associates), were the most aggressive in threatening market order with price cuts. No, Acting was a role L.J. wanted to shed as soon as possible.
Remaining competitive required product upgrades and new products. In February 1987, Proteon announced its p4210 gateway interface.86 Specifically designed to work with the p4200, the p4210 enabled its various token ring networks as well as those of IBM, including those of competitors like 3Com and Ungermann-Bass – any products conforming to the 802.5 standard—to communicate with, and route over, Ethernet networks.
In March 1987, Proteon announced a series of new enhancements for its Pro-NET-4 network to make the network compatible with: IBM’s PC LAN Program, Advanced Program-to-Program Communications PC, Network Basic I/O System and the 3270 Emulation Program Version 3.”87 Then in July, they announced more enhancements necessitated by competing with IBM, this time for its ProNET-10 networks.88 To match the price reductions of Ethernet competitors, Proteon also announced a 30% price reduction for its ProNET-10 adapters. Competition and lower sales prices were impacting revenues and margins, and higher operating costs depressed earnings. It was obvious to Sevin that more cash was needed both to protect the balance sheet as well as attract a candidate for President. At the August Board meeting, Sevin reported the search had uncovered no qualified candidates, and that the company’s distribution strategy needed to include many more value-added resellers (VARs). At the October meeting, the Board dealt largely with the yet-to-close financing. After a lengthy discussion of pricing, it seemed fair to price the round down by 30% from the last round – but not nearly as deep a “haircut” as some new investors wanted. Proteon had clearly lost its patina and was now seen more as a turnaround.
In November came the good news (the “financeable event”) the Board and existing investors had been waiting for: Patrick Courtin was named President, Chief Executive and member of the Board. He had the technology know how long sought, including ten years of experience with DEC. While his coming aboard had little impact on the financial results for 1987, he did facilitate closing the financing of $2.4 million, raising the total venture capital investment to $17.4 million.89 In December, he announced the projections for 1988: sales of $37 million and net income of $1.2 million; projections that helped in the closing the Preferred D round of financing. A month later the results for 1987 were made known: sales of $22.3 million, up 15% over 1986, with a loss of $5.7 million (a loss equal to one third of the venture capitalists’ total investment). Hopefully the yearend actions and decisions would contribute to a successful 1988.
At the Interface Trade Show in March 1988, Proteon introduced its second-generation p4200 gateway. Faster and more reliable, the new p4200 supported TCP/IP, DECnet and XNS, no longer was it just hardware interfaces but communication protocols as well. In an interview in May, Courtin said sales for 1988 were doing great, helped by the recently announced p4200. In fact, the sales of gateways and high-speed products, i.e. ProNET-80, accounted for 40% of Proteon’s sales.90
At the Board meeting in June, Courtin indicated sales for 1988 might total $40 million and then soar to $75 million in 1989. The sales breakdown for 1989 was estimated to be $50 million of adapters, $20 million of routers, i.e. gateways, and $5 million of high-speed products.
In November 1988, Proteon introduced a new router for smaller networks: it didn’t have a T-1 connection, but rather slower 64Kbps connections. Reflecting the demand for solutions, and a more complete product for it is corporate friend, Novell, the p4100 employed a TCP/IP backbone which connected to IPX(N), the Novell interface and communication protocol.
Proteon had a successful 1988. Sales of $40.6 million, up 82%, and income from operations of $1.1 million. Cash totaled $1.6 million. Proteon was second only to IBM in the market for token-ring adapters and a perceived leader in the fast-growing Internetworking market. If Proteon qualified as a turnaround, it might be said “so far so good.”
John Dix “Router man,” Network World March 27, 2006 http://www.networkworld.com/supp/2006/anniversary/032706-routerman.html
“Token ring ties MIT Ethernet local nets,” Network World May 1, 1989, p.27-30
Proteon ProNET Local Area Networks, pgs C11-695-101:106, Networks and Architectures, DataPro, Sept. 1986
IBM Token-Ring8 Network, pgs. C11-491-501:509, Networks and Architectures, DataPro Jan. 1986
“Interface Links Ethernet, Token Ring Networks,” Network World, Feb. 2, 1987, pp.- 22-23
“Proteon net to support IBM PS/2.! Network World, July 20, 1987, p.3
Proteon IPO documents
“Proteon chief pulls firm back on track,”, Network World May 16, 1988, pp. 25-26