Rewriting the History of Computing

by Albert Cory and Jerry Morrison

Rewind

In Silicon Valley folklore and legend, Xerox PARC invented almost all the important parts of modern computing, but other companies reaped the profits. Apple, Adobe, 3Com were heirs directly; almost everyone else, indirectly. As a page at Stanford says:

The closest thing in the history of computing to a Prometheus myth is the late 1979 visit to Xerox PARC by a group of Apple engineers and executives led by Steve Jobs. According to early reports, it was on this visit that Jobs discovered the mouse, windows, icons, and other technologies that had been developed at PARC… The Apple engineers-- that band of brothers, that bunch of pirates-- stole the fire of the gods, and gave it to the people.

Here’s that bunch of pirates. (Just think: that baby is now in his or her 40’s!)

The Xerox Thieves: Steve Jobs & Bill Gates

But then it goes on:

It's a good story. Unfortunately, it's also wrong in almost every way a story can be wrong.

Jerry and I were at Xerox then. We didn’t even hear of that “famous” visit until months later. That doesn’t mean anything, of course; lots of things only matter in hindsight. Were Xerox executives really as stupid as they look in hindsight? Xerox did get to purchase 100,000 shares of Apple, which wasn’t public at the time:

The Xerox Thieves: Steve Jobs & Bill Gates

What should they have done? Let’s go back to the beginning and rewrite the story.

Albert’s Take

“Albert Cory” is the pen name of Bob Purvy.

The problem was that Xerox wanted the Office of the Future to be a market they could dominate with the corporate structure they had. After all, that structure had let them own the copier market for twenty years and made “xerox” a verb

The Xerox Thieves: Steve Jobs & Bill Gates

In the “Meet Harold” chapter of Inventing the Future, which was based on a real meeting that we had with Don Massaro, our new leader in Dallas, Harold encapsulates Xerox’s attitude toward the new personal computer market:

“The office market is our market. We need this Xerox 820 personal computer to hold our place on the desktop, until the Star is ready.”

This really happened! The 820 was internally called “The Apple Killer.” [editor’s note: insert emojis here.]

Albert’s solution is exactly what Xerox didn’t want to do: unchain the new computing business from the old copier business. If Xerox doesn’t have the right sales force to push the new boxes, start a new company which will.

No Feelings of Entitlement

The comment “the office desktop: that’s our market” seems especially ludicrous, knowing how it turned out. To Xerox executives, the fact that these Apple users were bringing their little toy computers to the office just to run VisiCalc was a joke. Xerox was a giant corporation whose sales force knew every important manager in the Fortune 500! Who was Apple? Their current success must just be temporary.

Does a Silicon Valley startup act this way? Of course not—they know they’re always one step from bankruptcy.

Get Rid of Mid-level Management Involvement

Star’s development is riddled with this. When the Office Information Systems (OIS) project was just getting started, Dave Liddle, our head of SDD, “earned his stripes” (his words) in endless battles with Xerox management. You see it happening constantly in the book, where Grant Avery, a stand-in for all the people in “Planning” and other bureaucratic euphemisms, has a full-time job keeping corporate management happy. He is a fictional character, of course, but there were many like him in real life.

I have two stories about this; the first is real, the second may be mostly apocryphal.

IBM and the PC

IBM was a conservative company with a strict “white shirt and conservative suit” dress code for its sales people. They weren’t all men, but most of them were.

This article tells the story about the IBM Personal Computer, a world-changing invention. Unlike Xerox with its computers, IBM succeeded, at least for a while. Here’s the key part:

The logical place to build a small computer was inside IBM's General Products Division, which focused on minicomputers and the successful typewriter business. But the division had no budget or people to allocate to another machine. IBM CEO Frank T. Cary decided to fund the PC's development out of his own budget. He turned to William “Bill" Lowe, who had given some thought to the design of such a machine. Lowe reported directly to Cary, bypassing IBM's complex product-development bureaucracy, which had grown massively during the creation of the System/360 and S/370. The normal process to get a new product to market took four or five years, but the incipient PC market was moving too quickly for that.

Cary asked Lowe to come back in several months with a plan for developing a machine within a year and to find 40 people from across IBM and relocate them to Boca Raton, Fla.

Note the important parts:

The General Products Division would have made the PC, but they had no budget. In any case, they probably hated the idea of making a machine to compete with their other products.
Furthermore, if Cary had given them the PC, they would have taken four years, at least, and it would have been as expensive as the Xerox Star ($16,600). Why? Because all the IBM product managers would have had a say in it.
Bill Lowe gathered his own team of people and moved them to Boca Raton, instead of taking over another building in Poughkeepsie. In the pre-Internet age, this made it very difficult for the rest of IBM to even know what they were doing, let alone interfere.
Cary gave them a hard deadline. Lowe was not allowed to say “no, we need four years!” He had to get the PC out within a year, and so he used off-the-shelf products.

This independence was eventually withdrawn and PC development was pulled back into IBM’s main development groups. IBM immediately set about trying to kill the competition in the PC market, with the PC Jr., with its infamous “chicklet keyboard” which threw away one big thing IBM was still respected for: its ability, since the Selectric typewriter, to make keys that felt “the way a key should feel

IBM’s efforts failed and Microsoft and Intel scooped up all the money.

Beretta (the apocryphal story)

A long time ago I heard this story from someone at Xerox, and the reader should be warned that it’s unconfirmed, and maybe not 100% true. Call it a “myth.”

Beretta is the Italian firearms manufacturer you probably know from their pistols. The story was that they were the oldest continuous manufacturing company in the world (true), 800 years old (probably false; it’s more like 500).

Image from Beretta, 1960

To survive 800 years, Beretta had to have seen multiple generations of weapons come and go. Surely they must have had staff who grew up with one generation and resisted the next? How did they prevent these people from strangling the new business?

This is where the myth comes in, if that’s what it is: when they made swords, firearms came along. They started up that division in a different city. Thus, the sword-makers were unable to stifle the firearms division.

Similarly, for each new generation of weapons, they started it up in a different city. Clever, huh?

In researching Beretta for this article, I was unable to confirm all this. As far as I can tell, they started up in the early 1500s with a contract for muskets, in a city known for its iron-working.

In this case, the myth is “truthy” like many myths that aren’t literally true: to properly nurture a “disruptive” technology (modern jargon, sorry), you have to isolate it from the people who are its enemies.

Push the Technology Just Enough

In the chapter “Grant Learns the Gospel” on hardware alternatives for Star (which is taken from a real SDD planning doc on bitsavers.org) Grant sees that Xerox was trying to do something that was impossible right then. They ended up using a “bit-slice” processor from AMD to get the performance they needed, and it had too little memory

By kind permission of D. E. White, WhitePubs Enterprises, Inc.

The Star still cost $16,000 and even that wasn’t useful, since you also needed a file server and a print server.

In the discussion with a hardware expert, they discuss off-the-shelf microprocessors. The conclusion was that they weren’t powerful enough. Yet only a few years later, IBM released the PC with an off-the-shelf Intel 8088, and Apple released the Lisa with a Motorola 68000. What changed?

There are two answers to that, and they both relate to technological progress:

The microprocessors got better and cheaper.
The Mesa virtual machine was a relic of the old days, when memory had to be conserved at all costs. Compiling Mesa directly to the native machine language of one particular chip and optimizing it was simply ruled out. Yet that would have been the answer.

#2 is an example of a choice that a startup would usually not make. The Mesa virtual machine was a sacred cow within Xerox. The notion that programmers would write a lot of Mesa and then the virtual machine would be tuned to minimize memory usage was almost holy, and yet it prevented the company from doing what would have been the right thing eventually.

Even in the early 80s when using an off-the-shelf microprocessor was standard practice, one particular company, Gavilan, used an oddball programming language, Forth, purely because it was so memory-efficient.

Meanwhile, Microsoft Word, led by Charles Simonyi (ironically, a Xerox PARC alumnus), was written in C and was adequately fast on the hardware of the time.

Get Exposure to the Market

Almost five years went by from the inception of the Office Information Systems program at Xerox to the first commercial product. Only a very large company would ever do that. If you’re in a startup and raising money from venture capitalists, they want to see your idea validated, early, and with real money from real customers. They never want to wait five years to see their money at work.

I’ve heard endless arguments that Xerox could not have sold the Alto (too expensive, not suited to mass manufacturing, some IP may have been repurposed from the Data General Nova). No startup would have stood for that. To all those objections, the management would have just said, “OK, those are obstacles. Find a way around them.”

A commercial Alto would have cost a king’s ransom in 1977. Nonetheless, there was a market, albeit a very small one. Identifying those customers, selling it to them, and listening to their feedback was the small-business thing to do. There is nothing like the reality check of a real customer who offers more money if you satisfy them and no more money if you don’t.

If you say “this was too niche a business for a company as big as Xerox,” then you just proved the point. Giant companies don’t do that well. In this picture, “Housing funds in Argentina” on the bottom is a niche. Xerox only wanted to address markets in the second row.

YouTube, “What is a Niche Market? Definition and Examples”

Be Open

A mistake that is almost impossible to fathom, knowing how the market turned out, was that Star did not allow for third-party programs. Something like VisiCalc, which Apple didn’t provide for the Apple II, could never have emerged in such a system. No one company can foresee everything a customer would ever want.

I’ve heard it said that there was no third-party software industry in the late 70s. There are two counters to this:

Yes, there was. It was just much smaller. Examples: SyncSort, a utility found in many IBM mainframe shops, was one. Total, a database, was another, as was Adabas. Minicomputer companies like DEC, HP, and Data General, were heavily dependent on third-party software, like MUMPS. There was Whitesmith’s, a British company which sold a C compiler for under $1,000, and had, by their estimate, at least 225 licenses by 1980.
Even on the Apple II, the value of openness was impossible to miss, both in hardware and in software. Take VisiCalc, for the most obvious example. The computer magazines’ were full of ads for additions to your computer, although software hadn’t become a huge business yet (that’s why many of those ads were in the back!)

How Do We Do That?

My answer is not a simple Band-Aid, like “sell the Alto” or “be open.” In my view, the giant corporate bureaucracy would have screwed it up, however hard they had to work at it. IBM showed how a giant company could succeed at starting a new industry but still manage to lose it eventually.

Of course Xerox management wrestled with this question at the time, and we have it from a well-known researcher that “we don’t want to be a holding company” was their answer.

This same person doubted that Xerox executives were up to running a holding company, since that’s no easier than a copier company. He’s probably right about that, so this rewrite of history has to assume a massive upgrade of executive talent, and early retirements for those not up to the new tasks.

The question to answer is what does ‘holding company’ mean, and is that even the right term? Let’s look at some possibilities:

Corporate Conglomerate

Is this what we mean by “conglomerate”? Warren Buffet’s Berkshire Hathaway is a famous example. The next time you see the Geico Gecko on TV, think of Warren.

And we would be remiss not to mention that Xerox did try to go that route in 1982, when they bought the insurance company Crum & Forster, with the claim that this would “smooth out” the company’s earnings.

It’s difficult to pull this off. Not everyone can be Warren Buffett.

Spin-out

In a spin-out, a company divests itself of a business. A famous one was Hewlett-Packard spinning out Agilent in the 1990s. McDonald’s spun out Chipotle Mexican Grill in 2006.

Chipotle was a perfect example of the big company-little company culture class. CNN has a great story about that. Chipotle took McDonald’s money for eight years, opened lots of new locations, but generally rejected most of their big company ideas.

"Bless their hearts, McDonald's had a lot of great suggestions, and we were always polite about it," said Gretchen Selfridge, Chipotle's COO. "They really wanted us to do drive-throughs. They really wanted us to do breakfast. But we just really didn't do any of that."

The same clashes would have been true of a small tech startup within Xerox. But it would have to be able to reject the big company’s ideas when they were stupid.

Spin outs could be seen, cynically, as the flip side of conglomeration. Large company X acquires small company Y, discovers in a couple years that Y is growing much faster than X and has a better public reputation, and spins it out again. Its shareholders now have some Y stock as well as their X stock, and they’re happy.

Or the opposite: Y is a failure and is dragging down X’s stock. X spins it off and its stock rises because it no longer has the millstone of Y about its neck. X’s stockholders are again happy.

Spin-along

The various forms of “spinning” are a fruitful topic for business school researchers. Xerox’s problems were not unique, although for better or worse, PARC is probably the most famous example.

Here is a paper by a couple of Deutsche Telekom researchers, which describes another approach they dub the “spin-along.” What should be amusing is Table 1 and the problems experienced by large companies attempting radical innovation:

Do any of these sound familiar?

In a “spin-along” system, the parent company spins out an R&D project as a separate company, with the right of first refusal to spin it back in. The authors claim that Cisco does this, and Deutsche Telekom began such a program.

Contrary to the authors’ claims, Cisco did very few spin-ins, as I verified with a contact who was formerly an exec with them.

The authors note the problems with such a scheme. There is jealousy in the employees still in the parent company, and disappointment among the spun-back-in employees (who were important executives with some autonomy in the spun-out unit, but now they’re just mid-level executives again).

If the spun-out company fails, presumably the employees are allowed to rejoin the parent, chastened but wiser. If it does well, though, the parent company benefits from that success, much greater than the parent could have achieved on its own.

What PARC Actually Did Later

Finally, if we’re trying to figure out how PARC’s inventions should have been commercialized, we shouldn’t neglect how they actually did it after the early days. We’ll look at a couple of examples, where PARC launched a startup to commercialize its technology. In “Graceful exits and missed opportunities: Xerox's management of its technology spin-off organizations." (Henry Chesbrough, Business History Review 76.4 (2002)) the author reviews many of the companies that PARC allowed to leave Xerox.

Chesbrough notes that there was substantial resistance within Xerox to most of these startups. Bill Spencer, who took over PARC in 1983, had a rule of “no additional capital” which helped overcome the resistance. In most cases, a researcher could leave Xerox and receive a license to the technology that he or she had helped develop at PARC and perhaps some of their equipment, in return for which Xerox got an equity stake in the company, but contributed no additional funds. On the plus side, the founders had to attract outside investors; on the minus side, it limited the potential gain to Xerox.

Spectra Diode Labs

Spectra Diode Labs (SDL) was established as a joint venture of Spectra-Physics and Xerox, in April 1983. SDL was a result of laser diode research by PARC researchers including Donald Scifres, Robert Burnham, and William Striefer. Xerox and Spectra-Physics each had 40% of the company, with 20% for employees. Xerox put up no additional capital.

This was launched fairly close in time to the Xerox Star, so it’s interesting to ask “what was unique about it?” For one thing, Xerox’s need for laser diodes could be met by just buying them from a supplier, and those needs alone were not enough to support the very large scale production that would bring prices down. There was no question about whether owning laser diodes was “strategic” to Xerox.

To quote Chesbrough :

In 1992, Xerox realized a return of $15 million when the SDL management, led by Scifres, performed a cash buyout of both Spectra-Physics (itself now owned by Ciba-Geigy) and Xerox’s shares for $30.3 million. Three years later, in 1995, SDL went public at a market capitalization of $50 million. Recently, SDL experienced a tremendous surge in its growth, as its technology was applied beyond printers to high-speed optical communication switches. This proved to be a fortuitous new application for its technology, and the company was acquired by JDS Uniphase. The transaction was announced at an astounding $41.3 billion in June of 2000, though this value declined to $13 billion by the time the transaction closed in February of 2001. This amount again exceeded the total market value of Xerox in February of 2001. [emphasis added]

In other words, it was a big win for Xerox. It could have been much bigger.

Synoptics

Synoptcs was founded in 1985 by a couple of PARC managers, Andy Ludwick and Ron Schmidt, intending to use a technology called AstraNet for Ethernet over optical fibers. As Chesbrough relates it:

SynOptics was another PARC spin-off incorporated by two Xerox managers, Andy Ludwick and Ron Schmidt, in October 1985. Ludwick had come to Xerox from SDS in 1969 and had served in a variety of mid level management and sales positions. At PARC, Schmidt had developed an Ethernet system, then called AstraNet, that could operate on an optical cabling system. This would allow much faster transmission of network data than that offered by the coaxial cabling, which was the standard cabling at the time. For nearly two years the pair tried to sell the technology for adoption within an existing Xerox division. They encountered three key impediments, as Ludwick explained:

[First] none of Xerox’s operating divisions was interested in AstraNet because they did not have the right sales force to sell this system…. [Second], the divisions did not believe that it was possible to commercialize AstraNet within the limited window of opportunity that existed. They viewed everything in terms of copier machines, and it usually took four years to roll-out a new copier. Finally, since I thought that the market for AstraNet was at most $100 million, no one at Xerox wanted to invest any time in the technology.

Once on their own, Ludwick and Schmidt realized that the real market opportunity was not optical fiber but unshielded twisted-pair (UTP) wiring. UTP wiring was what was commonly found in offices, and customers loved not needing to run coaxial cable. Twisted pair wiring is almost universal now and “Ethernet” is the only game in town. You can buy one of their switches on eBay:

(At the time, I was at 3Com and encountered the Synoptics Ethernet switch, and I always wondered why it had “optics” in the name! Now I know.)

Synoptics took off. Within five years, its revenue was over $100 million, and it went public in 1988. It merged with Wellfleet to become Bay Networks, and is now part of Nortel. Clearly Xerox made some money from its stake in Synoptics, depending on when they sold. Had they actually invested capital beyond giving it a license, obviously they’d have made much more.

The Conclusion

With the Star, plus the “ones that got away” (3Com, Adobe, Apple) and the later ones that didn’t entirely get away (Spectra-Diode Labs, Synoptics, PlaceWare, and many others I didn’t write about), Xerox’s management viewed all new businesses through the lens of “strategic” or “not strategic.” A “strategic” business was related to copies or printers, period. Star was also thought to be strategic, but we don’t need to go over that again.

Even after they realized that letting researchers walk out the door with PARC technology didn’t make any sense and began taking equity in the startups, it was still because “this business isn’t strategic.” And there is also the quote: “We don’t want to be a holding company.”

So we ask, “Why not? Maybe Xerox’s real business wasn’t making copiers and printers; it was making money.”

Here is a rude question that would have gotten you thrown out of a Xerox boardroom: what did the shareholders deserve? To stay invested in what proved to be a declining business, or to get a return for their money? Did they have a sentimental attachment to the copier business? As I said: rude questions.

In my second book, I quote an engineer whose favorite saying was:

“Strategic” means you don’t make any money.

In this thought experiment, you’re allowed to just move Xerox management out of the way. So Albert’s solution is:

Dedicate the company to taking large minority stakes in many startups, all of which are required to find outside investors. Become like Berkshire-Hathaway (except they invest in large, proven businesses). If eventually the sum of those becomes larger than the copier business, then divest the copier business.

Jerry’s Take

PARC invented a lot of computing technology and made it practical, but translating research into products is always hit-and-miss. Over years, frustrated employees left to create new companies or new business lines within other companies.

Trying to make new products fit the copier company actually made a very tall order even less viable:

The “paperless office” vision meant not just obsoleting the copier business over time, it directly threatened the sales team’s commissions on paper and toner sales (the “razor blades” of their business).
A “turn-key office system” implied designing a closed computing product containing all the features that were obvious at the time -- so no third party applications and an unachievable to-do list for the initial release. Imagine trying to implement: word processing, embedded diagrams and charts, all the world’s languages, undo, email, networking, file and print servers, contacts, database records with mail merge, terminal emulators, … for a world that hadn’t seen italic text on screen.
Looking for “fitting” product ideas meant ignoring all other opportunities, like Remedy Corporation’s Action Request System.

Business Recommendations

First, don’t rely on a single product category. Nobody can predict what new products will succeed and when the timing will be right. Instead, help those frustrated employees start new companies. Invest seed funding and technology licenses. Help them get started quicker, and connect them to resources. Provide temporary office space and services like HR, legal, security, secretarial, and so on.

Second, get cracking on laser printers of all sizes. To be fair, Xerox did do that, starting with

the 9700:

Switch their copier technology to toner cartridges even if that required cross-licensing with Canon. (Around 1983 one could buy a small Xerox copier. Adding a service contract doubled the price! Or buy a Canon copier without a service contract knowing that nearly any failure would be in the replaceable toner cartridge.)

Third, consider reentering the computer business with GUI workstations. Xerox not only invented GUI workstations, they created three industry-leading programming environments: Interlisp-D, Smalltalk, and Mesa/Tajo. Xerox did eventually sell Interlisp-D workstations and licensed Smalltalk to Tektronix for its workstations. Such computers did not have the tight price constraints or colossal software expectations of an office system and might’ve fit the company’s direct sales team.

Project Lessons

Even though I don’t think the “office of the future” products could’ve succeeded at Xerox, I wish to send some kick-in-the-pants project management perspectives back in time to the Star project’s beginnings.

The first perspective is in Tom Gilb’s article, Evolutionary Delivery vs. the “Waterfall Model” in ACM SIGSOFT Software Engineering Notes, July 1985. (Also see his 1988 book, Principles of Software Engineering Man

agement.)

Evolutionary Delivery vs. the Waterfall Model: Don’t forecast how much you can do within a budget. The result always comes out late, over budget, and misses the mark. Instead seek the smallest thing useful (nowdays called a minimum viable product), deliver it to someone, get feedback, and evolve.

You need small teams, small steps, and empirical observations at every step to survive development complexities and real world unknowns. Even without Gilb's article, observe that successful software projects (including the Alto applications) were built incrementally by small teams, not using the forecast-the-future, guess-the-product-requirements, plan-it-all-out, implement-everything-then-debug-then-try-it-on-people, big-team waterfall model that Xerox adopted for follow-on copier products.

More project management clues come from Fred Brooks' 1975 book, The Mythical Man-Month. Here’s a famous quote:

In most projects, the first system built is barely usable… Hence plan to throw one away; you will, anyhow.

One of Apple’s strengths was their willingness to try again, such as the Lisa then Macintosh 128K then the successful Macintosh 512K.

Although a single do-it-all application might’ve been easier for end users, the variety of Alto applications in 1977 and then VisiCalc in 1979 should’ve made it clear that office systems need third party application programs.

The next perspective was in Geoffrey Moore’s 1991 book, Crossing the Chasm which built upon earlier research on early adopters.

You won’t succeed in creating a new product category -- let alone a new industry -- with a product that does 80% of what several customers need but doesn’t fully meet any one customer’s needs. To “cross the chasm” from early adopters to pragmatic customers, you must satisfy specific customer groups one by one.

Product innovation is timing-sensitive. If you’re too late, a competitor will get there first. If you’re too early, you might burn through your funding, find the underlying technology is too expensive, and discover that managers won’t touch a keyboard while secretaries don't get paid enough to put $47K workstations (in 2020 dollars) on their desks.