The Network is the Computer
The story of Sun Microsystems and the Java programming language
Andreas Maria Maximilian Freiherr von Mauchenheim genannt Bechtolsheim (or just Andy Bechtolsheim) was born on the 30th of September in 1955 at Hängeberg am Ammersee in Finning, Germany. His career in computing technology started at quite a young age. At 16, Bechtolsheim designed an industrial controller for a company that was based on the 8008. Being young, it isn’t as if he had a ton of money for fancy assemblers or compilers, so he wrote the necessary code for this controller in binary. The proceeds from this early career success would be put to good use in his education at the Technical University of Munich, and then at Carnegie Mellon University, and finally at Stanford. He earned his Master’s in computer science (1976), and then a Ph.D in electrical engineering (1978).
While working on his Ph.D at Stanford, Bechtolsheim spent time at Xerox PARC as unpaid consultant (essentially an intern). This gave him access to the rather legendary Alto, which he put to use in testing chip design tools. For Bechtolsheim, this experience showed him that there was a market for workstation computers for engineers and scientists. The Alto wasn’t turning into a commercially viable product, and therefore Bechtolsheim knew he’d have to design the machine himself. This project was initially funded by DARPA, and it involved several other individuals at Stanford. Bechtolsheim called this the Sun Workstation. Licenses for the design were sold at $10,000 in 1980 ($35516.38 in 2023). Sometime in 1981, Vinod Khosla met Bechtolsheim. Khosla liked the design, and Bechtolsheim offered him a license at the standard rate. Then something happened. Khosla said:
I want you, not your technology. I don’t want the golden egg, I want the goose.
These two men were joined by William Joy and Scott McNealy, and on the 24th of February in 1982, they founded Sun Microsystems. All of these men are Stanford graduates (except for Joy who went to Berkley), and the name “Sun” is derived from Stanford University Network. This is well named as from the start, Sun systems included network capability. Employee 5, John Gage, went so far as to say “the network is the computer,” which became the Sun slogan. Funding for this adventure was provided by Eastman Kodak, AT&T, Olivetti, and Xerox.
The Sun-1 workstation used a Motorola 68000 CPU at 10MHz. This was paired with an in-house designed MMU. It had 256K zero wait state memory with parity, and 32K EPROM memory. The machine sported two serial ports, one 16 bit parallel port, and an IEEE 796 bus interface. The first serial port (A) was wired as a DCE port (modem), while the second (B) was a DTE port (no modem control). The parallel port allowed the connection of an 8 bit keyboard and 8 bit mouse. It could not be used for printers. Video output was provided via a frame buffer board that had a resolution of 1024x1024 with 1024x800 displayed on screen. If a machine were to use color that resolution dropped to 640x512 with 640x480 on screen. In this case, color meant 256 on-screen from a palette of 16 million. Memory expansion was possible with up to two memory cards for a maximum of 2MB. While these memory cards were multibus form factor cards, that bus only provided the power. Memory cards used the P2 bus for private, synchronous, zero wait state memory access. This machine could also make use of an ethernet board. The first ethernet board available for the Sun-1 implemented the Xerox PARC 3Mbit/s spec. A later board would provide the 3Com 10Mbit/s spec. There were also disk controllers and tape controllers available for the machine. The first Sun-1 machines sold were actually used as terminals for an IBM 360. The rest, however, ran 4.2BSD (Bill Joy was a developer of BSD). The Sun-1 sold for $8900 in 1982 ($26,990.98 in 2023). These early machines were followed by the Sun-2 series which would refine and improve the design. On the software side, SunOS (derived from 4.2BSD) would see improvements like NFS (which Sun invented), RPC, and VFS.
1986 was a big year for Sun. The company went public (with symbol SUNW), they developed the SPARC CPU, and they sold 500 million dollars worth of equipment to NSA ($1516347150.26 in 2023). By the end of the year, Sun hit 1 billion dollars in sales. The SPARC design was inspired by Berkley RISC, and it was first used in Sun servers. In 1987, there was some discussion internally at Sun about expanding the use of the chip into workstations. This was viewed as quite risky. Sun was newly public and investors would be expecting conservative moves. Bechtolsheim, on the other hand, was concerned about competition. Steve Jobs was building NeXT and his machine would use the 68000 in a workstation, and Bechtolsheim wanted to have the better product.
By this point Khosla had left Sun and joined a venture capital firm, and he provided the funding for a new Bechtolsheim venture, Unisun. At Unisun, the new Sun workstation would take shape around the SPARC CPU. This technology was then brought back into Sun when it proved to be far more powerful and capable than the 68000 workstations. Sun shipped the Sun-4 line (officially called the SPARCstation) in 1989, and all of these machines’ SPARC CPUs clocked between 14.28MHz and 33MHz. These machines were based around a VME bus like the Sun-3 servers (officially called the SPARCserver) that preceded them. The SPARCstation was a major success for Sun.
Sun was growing faster than any other US company. Between 1985 and 1989, the compound annual growth rate was 145%. Sun was the undisputed king of the workstation market.
In the 1990s, computing began to change from small networks and standalone desktop machines to very large networks and internet-connected computing devices of all sizes. Patrick Naughton had no intention of missing out on the future of computing. In his eyes, Sun was a hot mess. He wanted out, and Jobs’ NeXT Computer was his intended destination. Before he left, however, he was having some beers with James Gosling. These two were on the same ice hockey team. Naughton spoke about his intention to leave, but he also mentioned why. In his view, NeXT was doing better stuff. Gosling then asked him to write out not only his complaints about Sun but also his solutions to the perceived problems before he left the company. Naughton obliged.
What he wrote wasn’t a simple email. This was a 12 page report comparing NeXT and Sun, and it detailed the strengths and weaknesses of the two competing workstations. Naughton argued that Sun needed to focus on a single programming language, a single GUI toolkit, and a single windowing system. The report lit a firestorm within Sun. Hundreds of people responded in agreement, and the management team knew they needed to do something.
John Gage (director of Sun’s science office) asked Naughton what he really wanted to do in a small meeting (once described as a “b&%$fest”) with a group of engineers and a few executives that ran from afternoon until past 4AM the following morning. With Gage’s support, Naughton took the idea to Wayne Rosing (President of Sun Laboratories). The idea was to have small group work on smaller, more personal systems. Naughton also wanted this team to be rather small and physically removed from the rest of the company. It was a requirement from Naughton that only the highest level executives know about this group and its mission, and also that the team wouldn’t be required to make any of this new technology compatible with other Sun hardware and software. For this, Naughton requested $1 million in funding for the first year.
Rosing took the proposal to McNealy over dinner, and things went well. Following dinner, Rosing called his assistant by car phone. His assistant emailed Naughton and said that McNealy expected full approval. Bill Joy and Andy Bechtolsheim both approved, and Naughton got everything he wanted.
Patrick Naughton, James Gosling, and Mike Sheridan formed a group codenamed “Green.” They wanted to build compact and simple computing devices that blended into life. In the words of Naughton in a Wired Magazine interview:
We wanted computers to go away, to instead become an everyday thing. We thought the third wave of computing would be driven by consumer electronics. The hardware would come from Circuit City, and the software would come from Tower Records.
In the Spring of 1991, Sun held its annual retreat for high-level staff. The Green Team was included, and the team had expanded to include hardware engineer Ed Frank. After some beer and while soaking in a hot tub, the four discussed how ubiquitous computing machinery had become. Chips were now in everything… including doorknobs at the resort, but none of these devices were programmable, and none of them could communicate with one another. So, lubricated with alcohol and soaked in hot water, the team decided to build a prototype device that could interact with all of these devices.
The team moved to an office above the Bank of America on San Hill Road in Menlo Park. They then proceeded to drink Coca Cola, play some Nintendo, eat some ice cream, and let their minds wander. Wander they did. Their mission statement was titled “Behind and the Green Door,” and it read:
To develop and license an operating environment for consumer devices that enables services and information to be persuasively presented via the emerging digital infrastructure.
At a concert, Gosling had a realization that he would need a new language. It would need to have both native networking capabilities and GUI functionality. He wanted the language to have syntax similar to C and C++ for the ease of programmers who were already familiar with those languages. The final realization was that to make this language portable to all of these new myriad devices, he’d need it to use a VM. Gosling already had experience with this type of work as he’d previously ported PERQ Q-Code to VAX assembly and had built a VM to emulate the PERQ hardware. This new language was called “Oak” after a tree outside the office window. From there, the design was boiled down to 5 principles.
It must be simple, object-oriented, and familiar.
It must be robust and secure.
It must be architecture-neutral and portable.
It must execute with high performance.
It must be interpreted, threaded, and dynamic.
Using Oak to program it, the team then proceeded to tear apart a variety of appliances and gadgets and from those assemble the device of which they’d dreamed. It was a box roughly the size of a softball with internal batteries and a touch screen on one side. There were no buttons. Interaction took place by tapping or sliding a finger on the touch screen. There were also no menus. You could move an “agent” (this character became the Java mascot, Duke) through a virtual world, and pick up a TV guide, select your program, and slide the program title onto a virtual VCR. The device in your hands would now program the VCR to record the show. The device was called *7 (star seven). This was before both Magic Cap and Microsoft Bob. The executives at Sun loved it.
The Green Team had been reaching out to megacorporations and pitching the technology, and Sun setup a wholly owned subsidiary company called FirstPerson Inc. At its founding this company consisted of: Lisa Friendly, James Gosling, Jonni Kanerva, Tim Lindholm, Patrick Naughton, Kim Polese, Lisa Poulson, Wayne Rosing, Eric Schmidt, and Mike Sheridan. The team expanded, and Sheridan left when Rosing was promoted to President of FirstPerson. They got new offices in down town Palo Alto. Unfortunately, no one was really interested in *7.
The information superhighway was all the rage, and many people were thinking that interactive television would be the main route into the home. After all, people already had cable and televisions, so this would be a somewhat natural move. Time Warner circulated a proposal to start interactive television trials in Orlando in March of 1993. To this end, the team pivoted and built a prototype smart settop box to link the TV to the information superhighway using Oak. This included graphics, images, sound, video, and payments. Ultimately, they lost. SGI got that deal.
The technology then made its way over to 3DO who wanted their game console to double as a smart settop box. The problem was that 3DO wanted exclusive rights to that technology, and FirstPerson walked away.
FirstPerson was falling apart and the executives at Sun were demanding profits. In March of 1994, Sun closed FirstPerson.
Bill Joy, however, saw the promise of Oak. The World Wide Web was here, and Joy figured that Oak should be put to use in that arena. He recruited Gosling to work on adapting Oak itself, and he recruited Naughton to write a killer app. In December of 1994, these were leaked onto the web and a link was given to Marc Andreessen. He loved it, and he told San Jose Mercury News:
What these guys are doing is undeniably, absolutely new. It's great stuff.
In January of 1995, Oak was renamed to Java, partially for marketing reasons and partially due to a copyright already existing on the Oak name. The killer app was HotJava which was a web browser built in Java as well as a HotJava JavaBean component for HTML rendering in other Java applications. A big move was made by Joy and Gosling (Naughton left that year as he didn’t get the kind of raise for which he’d been hoping), they chose to make Java technology freely available over the web. Importantly, they also struck a deal with Netscape. It was seriously discounted, but Sun wanted Netscape to enable Java Applets within their browser to spread Java as quickly as possible. It worked.
At the same time that Java was getting started, Sun began working on another issue that Naughton had brought up, their OS. Work began on the 4th of September in 1991 to replace SunOS 4 based upon BSD with Solaris based upon UNIX SVR4. Solaris used the OpenWindows environment which supported both X Windows protocols and NeWS. However, Sun was working with IBM and HP to develop the Common Desktop Environment (CDE) which would be released in 1993. SunOS 4 was retro-actively rebranded as Solaris 1, and SunOS 5 would be shipped as Solaris 2. Solaris 2.1 shipped in December of 1993 supporting both SPARC and x86, and this version also had SMP support. Solaris would go on to innovate many different technologies, and it would do so quite early. It had SMP and multithreading before many other systems among many other technologies. Solaris was the birth place of ZFS and dtrace. Solaris had containers before most people in the industry knew what containers were. Solaris was also quick to move to GNOME and build upon it with the Java Desktop System. Sun’s StarOffice (later OpenOffice) debuted on Solaris. Like SunOS before it, it was one of the most important unices available.
In 1998, Sun’s revenues crossed $10 billion, and Sun’s net income stood at $763 million. Sun licensed Java to Sony, Motorola, Ericsson, Samsung, Alcatel, Nortel, OpenTV, BEA Systems, Siemens-Nixdorf, and Scientific Atlanta in 1999.
When Sun went public, it was trading at 77 cents. By late Summer in 2000, the stock price hit $257.25. The internet was big, and Sun was the internet in the eyes of many. Sun servers running Solaris and Java were seen as the premium server stack for all who could afford the price tag. Thanks to the dot com bubble, many could afford that price tag. Sun benefited.
ARM9EJ-S in 2001 included support for the efficient execution of Java bytecode. This made both Java and ARM the standard for mobile devices in a way analogous to Windows and Intel on desktop devices.
The dot com bubble burst. By December of 2001, Sun’s stock price had fallen to $100 and it was still falling. It would hit $10 at its post-bubble nadir (it would rise again later on, but then fall below that $10 price, never reaching far). A company built on networks and taken to staggering heights by the World Wide Web was going to pay a huge price when so many of its customers went out of business. By the 9th of October in 2002 the stock market generally had lost $5 trillion in market capitalization and the NASDAQ-100 had lost 78% of its peak value.
The market crash alone did not kill off Sun Microsystems, but it weakened the company considerably. Several executives left the company, layoffs ensued, and Sun killed off several projects and departments. Revenue was coming in, but losses were also felt. The rapid fall slowed to a steady decline.
SPARC didn’t keep its edge. The x86 platform became extremely competitive with the arrival of AMD64. The 64 bit x86 machines were cheaper than SPARC machines, and when coupled with Linux they could run all of the UNIX software any web company needed. This fact was proven by the likes of Google and Facebook. The core of Sun’s business was destroyed.
In 2009, Sun was sold to Oracle at a price of $5.6 billion. Quite a bit of staff was laid off, and the acquisition was completed in 2010. Sun’s campus in Menlo Park was bought by Facebook.
Java is still going strong. It enables both Android and Minecraft among many other products and services. ZFS is now a preferred filesystem. NFS has been ubiquitous in the server world for more than 20 years. Dtrace is still an awesome tool. OpenOffice was forked to become LibreOffice. Sun’s legacy is alive, and now more than ever before, the network is the computer.