Inside the Invention of the Stanford Router That Inspired Cisco

If you wanted to let two computers talk to each other, how would you do it? With just two machines, the answer is simple: connect them with a cable and let them pass information back and forth. But what if there were more than two computers? And what if each of those computers spoke in a different language?

Such was the situation on Stanford’s campus in the late 1970s and the 1980s. Early computing pioneers like Ed Feigenbaum and Vint Cerf worked with a panoply of early computers including Digital Equipment Corporation PDP-10s and PDP-11s, Xerox PARC Altos, Dolphins, Dandelions, and Dorados, Symbolics Lisp machines, and Texas Instruments Explorers carrying out seminal work in the fields of artificial intelligence, algorithm analysis, computer graphics, hardware, databases, and more. The machines were top tier for their day, but getting information between them was a hassle, often involving physically carrying magnetic media across campus. Even hooking the computers together using the newly invented Ethernet-based local area network technology was problematic because early Ethernets were made of multiple segments of Ethernet cables, each limited to a length of 2,000 feet. Further, each computer manufacturer used its own proprietary networking protocols.

“The business of being able to hook diverse kinds of computing systems together became really important to our AI research,” says Tom Rindfleisch, the director of Stanford’s Knowledge Systems Laboratory at the time. “Lacking commercial products, we decided to develop our own.”

Rindfleisch handed the task of creating a device capable of linking together Stanford’s disparate computer systems to a research software engineer named William Yeager, who was working with the pioneering Stanford University Medical EXperimentation in Artificial Intelligence and Medicine (SUMEX-AIM) project. Making a multiprotocol “gateway” – what we now call a router – that would let disparate computer systems communicate contained deceptive complexities. That suited Yeager just fine.

“My attitude was, and I've always had this attitude, if there's something that no one else wants to do and it's difficult, I will do it,” said Yeager, who retired from Stanford as a Computer Systems Specialist in 1994.

Getting From Here to There

Though the basic concept of routing packets of information between computers was not new, Yeager was starting almost from scratch when it came to building a router that could handle Stanford’s evolving and heterogeneous computing environment. That meant writing an entire operating system from the ground up and running rigorous tests to debug the system and confirm its capabilities. Adding to the challenge, the computing limitations of the time meant the router code had to be lean enough to run on the 56 kilobytes of memory available on the PDP-11/05. Yeager spent the summer of 1980 building and refining his software, including writing an optimizer for the C compiler object code to reduce the code base by a third.

Within three months, Yeager had spun up the operating system for his router and, with the help of hardware engineer Nick Veizades, built the physical router. The first multiprotocol router sent a maximum of 320 kilobits per second of data via 3 megabits per second (Mbps) Ethernet between the various computers on campus. Though painfully slow by today’s standards, it was a quantum leap in networking, providing a means of near-instantaneous communication between computing groups physically separated from each other.

It was also admirably well-built. Yeager recalls talking with Chuck Hedrick, a computer scientist at Rutgers University, one of the first sites outside of Stanford where a later version of the router was installed, who told him it was the first software he’d used that never crashed.

“[The router] was remarkably stable. He was a perfectionist coder,” Rindfleisch said. “It was a labor of love for him.”

Until Yeager, no one had created software that could route data packets for four distinct networking protocols, allowing those using different systems to seamlessly communicate. The project was responding directly to the needs of Stanford’s research computing environment, but the router nevertheless anticipated the fundamentals of diverse internet connectivity today – years before the world wide web and mobile computing existed. Yeager’s homegrown solution to the problem of directing information to where it needed to go laid the groundwork for the digital solutions we rely on daily, from internet searches to email to video calls.

But that would all come later. In 1980, Yeager’s first router was installed in a closet in Pine Hall, midway between the Stanford Medical Center and the Computer Science Department in Margaret Jacks Hall, connected by thousands of feet of copper-core coaxial cable. After building the system, Yeager volunteered to maintain the routers for Stanford’s local area network, which often meant late nights with Veizades crawling around to check on the cables and electronics for the SUMEX-AIM networks, and, later, monitoring the network from a computer terminal in his home nearby.

A Search for Perfection

As computer and networking technology progressed throughout the 1980s, Yeager kept pace. He rewrote the router and its operating system in a modern version of C in 1982 to use the new Motorola M68000-processor motherboard developed by Andy Bechtolsheim, a computer science graduate student who would go on to found Sun Microsystems. Other additions included 10 Mbps ethernet connectivity and routing to and from the internet in 1983 via Stanford’s “Golden Gateway.” Yeager remained the driving force and primary architect for the router, with Donald Knuth, Vaughan Pratt, and Jeff Mogul making small contributions, such as an efficient C macro and an algorithm for pseudo-random number generation. Throughout, Yeager obsessed over performance, streamlining his software to squeeze every byte per second from his network.

“My goal was to make the routers run as fast as the hardware could actually provide the speed to do,” he said. “All my stuff ran at bus rates,” or the maximum speed at which a processor could shuttle information back and forth internally.

Yeager’s perfectionist bent dovetailed fortuitously with the culture of open-ended invention that prevailed in Stanford’s Computer Science Department during that era. Faculty in the early days of computing at Stanford took on brand-new questions that had little commercial relevance at the time, but which would come to be hugely important for the growth of computing technology later, providing the basis for numerous world-conquering software companies of the 1990s through to today.

“Invention was a crucial part of everything that got done,” Rindfleisch said. “If you wanted to build a new technology, you had to start from the basics and pull this stuff together.”

It wasn’t long before Yeager’s router attracted the attention of Len Bosack and Sandy Lerner, co-founders of Cisco Systems, today a multibillion-dollar networking company. Their first product was a router inspired by Yeager’s software, which the company eventually licensed from Stanford in 1987 through the Office of Technology Licensing. Yeager refused to take the royalty money for himself, instead returning it to the Computer Science Department to be placed in a personal unrestricted fund to further his research. In a karmic twist, the money would later pay Yeager’s own salary for a time when his research funding was temporarily cut after the corporate partner ran into financial difficulties.

Yeager continued updating and maintaining his router software throughout the 1980s and into the 1990s before leaving in 1994 to work for Sun Microsystems. His later career accomplishments include work on the IMAP protocol still used for email today with Mark Crispin, a SUMEX-AIM staff member. At Sun his contributions included inventing the high performing Sun Internet Message Service, SIMS and helping lead Project JXTA, the first open-source peer-to-peer project.

Today, at 85, Yeager still plays tennis on the Stanford campus once a week and walks the Stanford hills as well as around his home with his wife, Joan. They celebrated their 59th wedding anniversary this past June. His contributions to computing history were recently memorialized by the Computer History Museum (CHM) in Mountain View, California, which published a detailed technical paper by Yeager describing the development of his routing software. The CHM also chronicled his life in a 2014 oral history project with Dag Spicer, the Museum's Senior Curator.

It is now a cornerstone of internet connectivity, but the multiprotocol router began simply as a research project, and a uniquely Stanford one. A mix of diverse computing tools and curious, passionate individuals forged a software solution that was ahead of its time – and one that was impeccably coded to boot. It’s a story that has played out time after time at Stanford, where open-ended experimentation becomes bright new ideas, which become innovative technologies that benefit the world at large. That process is enabled by resources like OTL, which ensures Stanford ideas are translated effectively and beneficially to society, while ensuring inventors like Yeager get their due.

For his part, Yeager remembers the time creating his router technology fondly, his memory little dimmed.

“It was a joyous experience,” he said.