Although, picking TCP/IP proved to be the right decision ultimately, it was any thing but easy decision at the time. TCP/IP was seen as tainted protocol, developed by DARPA ( Defense Advanced Research Projects Administration) for long haul, slow and unreliable networks using phone lines and was seen as unsuitable for high speed reliable networks like Ethernet . It had very little industry support. Xerox PARC had developed XNS as lean mean protocols to go along with Ethernet and most of the networking start-ups were adopting them. Most of the industry groups were adopting ISO protocols as their favored protocols. IBM and DEC had their proprietary networking protocols (SNA and DECnet) that they were selling and promising a future move to ISO. Big customers like Boeing and GM had lined up behind ISO too.
The problem we were faced with was very simple. How do we sell enough of hardware and software in a very short order to survive as a company. We knew customers have an immediate problem of connecting a plethora of computers they had bought. How to knit them in to a working network? Xerox had put XNS protocols in the public domain but had decided to hold back applications. Various start-ups were developing their own proprietary applications, essentially making them non-inter-operable with with others. ISO protocol were in early stage of development with no working implementations. IBM and DEC were too happy to pay lip service to ISO protocols while trying to lock-in customers in to their proprietary networks.
We figured that a working solution is the need of the moment. TCP/IP offered that. It had implementations on all sort of computers and operating systems. It had two very useful ready made applications: FTP (File Transfer Protocols) and Telnet (Virtual Terminal Protocols) that customers could use immediately. We also figured that though they were designed to work in slow and error prone networks where response times were slow and extensive recovery routines were needed to make network work reliably, there was nothing inherently wrong with protocols so they would have trouble on high speed reliable networks. The error recovery software would not burden the network if there were no errors.
Excelan was a technology supplier to OEMs when I took over. We had over 30 OEMs but none was taking the volume. I decided to go to the end-users directly by packaging solutions for IBM PCs, DEC VAXes and UNIX Machines and promoting them directly to the end users. The packages included every thing customer would need to connect his computer to the Ethernet networks. All the hardware, software, cables, nuts and bolts were included in the box. Customers were offered full satisfaction or money back warranties. We had priced in 10% dissatisfaction where we may have to fly out a technician at our expense to make the network work. Solution was an immediate hit. Only 1% of the customers needed help initially and it went down from there as we improved our documentation and process.
Initially, we thought our target market would be scientific computing market. This meant research labs, engineering departments and universities. That’s where we focused our marketing energy. We were surprised to see Wall Street emerging as a largest market. As a matter of fact we were not able to identify any one industry that preferred our solution, we were getting business from all sort of customers in all sort of industries. Networking was one of those horizontal capability that everybody needed and we were the only game in town if you had a heterogeneous environment.
We were feeling very good about our business and counted marquee names as our customers. But press and industry associations were forecasting an immediate demise of TCP/IP and mass adoption of OSI Protocols. Boeing and GM, both of them a very large customers of Excelan, joined hands to form an association to promote MAP/TOP (manufacturing automation protocols; Technical oriented protocols) built on OSI protocols. Most of the European companies lined-up behind this effort. Siemens, an OEM of ours, had forbidden us to claim them as a customer as they were a big champion of MAP/TOP. GM corporate networking Czar summoned me to come visit him in Detroit. I was told in no uncertain term that I had to drop TCP/IP and move to OSI protocols if we werw to keep GM as a customer. Excelan would be black listed soon and there would be no more business from GM unless we publicly announced that decision.
It was a strange quandary as our business kept getting stronger while disparaging of TCP/IP was reaching a crescendo by the fall of 1985. I was getting more convinced that TCP/IP would be an eventual winner as it was a working solution that was being adopted widely. While other protocols were being hyped, TCP/IP was improving and getting battle hardened under real use. It was hard for me to imagine that customers will rip out their working networks to install some unknown new protocols.
Among all this confusion, we got our largest PO (purchase order) from GM. I realized that corporate Czars have a luxury of indulging in their fantasies while operating managers need working solutions. It was a sure sure sign that time had changed when Siemens came back to ask for the relief from the agreement that neither Excelan and nor Siemens, could publicly announce the relationship.
It was another couple of years before TCP/IP was widely seen as the future of networking and another 10 years before the emergence of the browser and the Internet. Even though very lonely as the only champion of the TCP/IP in the industry, I felt secure in my decision to bet the company by those in coming POs!