Tech Nostalgia: A Computer Needs an OS Like a Fish Needs a Bicycle (Premium)

The first personal and home computers booted into a ROM, or read-only memory, that usually consisted solely of an on-chip BASIC interpreter, typically made by Microsoft. That was minimally acceptable for the software developers and other enthusiasts and early adopters who constituted the market for these devices in the 1970s. But for personal computing to enter the mainstream, these machines would need to become more sophisticated and more useful. And both of those attributes would be achieved largely through software, not hardware.

The more powerful mainframes and minicomputers of the day used so-called operating system (OS) software like Unix or TOPS-10 that ran at boot time and acted as an intermediary between the software programs that users ran and the underlying hardware. These OSes handled input and output, processor, RAM, and other resources, and provided common services so that programs–what we call apps today–didn’t have to recreate each individually. In that day, of course, user interfaces were minimal and entirely text-based, and so they were still relatively unfriendly and unsuitable for mainstream users.

By the late 1970s, Microsoft’s Bill Gates saw the need for a similar “standard OS” for personal computers. His company quickly found itself supporting multiple incompatible microprocessor architectures–from the Intel 808x series to the MOS 6502 to the Motorola 6800 and more–and a coming generation of 16-bit chips would provide even more powerful hardware for even more advanced computers. Standardization would help easy the difficulty and time it took to port programming languages and programs between these machines.

“Software houses would write programs to run under the standard OS and wouldn’t have to worry about multiple versions,” he wrote in his column in Personal Computing in 1977.

Today, it seems obvious that Microsoft would tackle this problem. But Gates had no plans to make an OS when he wrote those words. He didn’t see the need: Gary Kildall had already created an OS for 8-bit microcomputers called CP/M–Control Program/Monitor, later the slightly less terrible Control Program for Microcomputers–that focused on file storage. This so-called disk operating system, or DOS, originally targeted the same Intel 8080 processor that inspired the creation of Microsoft, but it quickly spread to other architectures like the Zilog Z80. Indeed, it was so popular with Intel- and Z80-based computers that Microsoft created its first-ever hardware product, the Z80 Soft Card, in 1980 so that users of the incompatible 6502-based Apple II line could access the then-vast CP/M software library.

As a DOS, CP/M managed file storage and organization, sparing app developers from having to write the code to do so themselves. It also provided a software environment in which programming languages more sophisticated than Microsoft’s interpreted BASIC could run, enabling them to create ever-more-powerful programs. Microsoft at the time was working on a version of the Fortran programming language that Gates realized would benefit from CP/M. And while Kildall had earlier created a BASIC compiler, he wasn’t interested in programming languages. And so it was only natural that the two would meet and discuss how Microsoft, and Kildall’s company, Digital Research–amusingly named Intergalactic Digital Research at first–might work together. In a perhaps too-perfect “only in California” moment, the two even discussed a merger in Kildall’s hot-tub during one meeting.

“At this time, I was quite friendly with Bill, and we always talked of the ‘synergism’ of our two businesses,” Kildall later wrote in a never-completed memoir in which he also discussed his distrust of Gates. “We talked of merging our companies in the Pacific Grove [California] area. and we both took that into consideration. Our conversations were friendly, but, for some reason, I have always felt uneasy around Bill. I always kept one hand on my wallet, and the other on my program listings. It was no different that day.”

This is one of the great what-ifs from the dawn of the personal computing era, and while Gates and Kildall were more different than they were similar, they had complementary skills, products, and goals. Microsoft co-founder Paul Allen argued to Gates that their company was better off moving to their hometown of Seattle–from Albuquerque, New Mexico, where their partnership was formed–than to the Bay Area “Silicon Valley” where Kildall ran Digital Research. And so that’s what Microsoft did, leaving the OS market, limited as it was by its size and architecture support, to Kildall’s CP/M.

“The combination of Kildall and Gates could have been a killer-deal in those days,” Kildall continued. “I had the operating systems for the decade to come, and he had the opportunistic approach to garner business. But, our attitudes differed entirely, and that could also have been a disaster. I think we both realized this and simply let the ‘deal’ die.”

But their paths would keep crossing. In addition to Microsoft’s Z80 SoftCard, which went on to sell with as many copies of CP/M as Digital Research sold itself, IBM entered the scene in the early 1980s with a need to build its first personal computer and to do so on a tight schedule. IBM had originally considered reselling an existing computer, and the Atari 800 was seriously considered until IBM chairman Frank Cary nixed the deal, curious why the most powerful technology firm on earth couldn’t make its own computer. So IBM started the Manhattan Project internally–yes, it really used that name–with the internal goal of having a small team working outside the normal hierarchy produce an all-new personal computer within one year.

To achieve the seemingly impossible, IBM would need to use off-the-shelf hardware and software. It adopted a hybrid Intel 16/8-bit processor called the 8088, a less expensive version than the fully 16-bit 8086 that was nonetheless dramatically more powerful than the 8-bit systems of the day. And it utilized an open architecture so that others could support the machine with add-in hardware cards, as with the Apple II.

On the software front, IBM approached Microsoft for BASIC and other programming languages, but it knew it needed an OS as well. And it was impressed enough with Gates and Microsoft–and had such a short schedule–that it asked if they had an OS that IBM could use as well. Perhaps it could sublicense the CP/M that Microsoft sold with its SoftCard?

Gates explained that the CP/M that Microsoft used was 8-bit software that wouldn’t work on the the IBM machine’s 16-bit processor. But his buddy Gary Kildall was working on a 16-bit version of CP/M, and so he personally arranged a meeting between Kildall and IBM to put the deal in motion.

What transpired next is one of the most widely known stories in personal computing, a Genesis story for the ages. Unfortunately, everyone seems to remember it incorrectly, or at least misunderstand what happened.

It’s not complicated.

Digital Research had been working on a 16-bit version of CP/M called CP/M-86 for two years, and it was still incomplete by the time IBM came calling. Not helping matters, a personality clash got in the way as well.

Where Bill Gates had pushed back a meeting with Atari CEO Ray Kassar by one day so he could meet with IBM instead, Kildall was a different person with different goals. He also had meetings scheduled for the day and time IBM wanted to meet, and he wouldn’t offend existing customers by canceling or delaying those meetings. As important, Kildall’s wife Dorothy McEwen handled Digital Research’s day-to-day business anyway, and so she was on hand to meet with IBM, as was always the case with potential partners or other business deals.

Gates had immediately signed IBM’s one-sided non-disclosure agreement (NDA) after barely reading it because he was so eager to make a deal with IBM. But the more experienced McEwen carefully read the NDA and refused to sign. A stalemate ensued. And then the stories about Gary Kildall being “off flying around”–he used a personal plane to get to and from far-off meetings–began. Long story short, IBM went back to Gates and Microsoft for help. Gates interacted with Kildall again, who was as confused as anyone about the events that had transpired. And while he was ready to make a deal, Digital Research lost out because CP/M-86 was still so far behind schedule. It just wasn’t ready, and IBM needed to move quickly. Too quickly.

Once again, Gates was bailed out by Paul Allen.

The other Microsoft co-founder knew of a local engineer named Tim Paterson who had worked up a CP/M-86 clone called 86-DOS. (The original name, and the one everyone remembers, was QDOS, for Quick and Dirty OS). 86-DOS ran on one of the world’s first Intel 8086-based computers and required just 6K of code. He had only created it out of frustration: He, too, had been waiting on CP/M-86, and he felt that just putting Microsoft BASIC on the computer–which his employer had made–was too limited. Allen suggested that perhaps Microsoft could license 86-DOS cheaply, improve it as needed, and license it to IBM. Doing so non-exclusively was Gates’ contribution, inspired by issues Microsoft had had with its first partner, MITS. It would go on to be the smartest business decision of his life.

Paterson was happy to make a deal, but he was unaware of its scope. Microsoft–referencing the NDA it had signed–never disclosed who the ultimate customer would be. He was later hired by Microsoft, where he worked on 86-DOS and then the SoftCard, and he’s never complained about the relatively modest sum he received for creating 86-DOS.

“If Digital Research had delivered, there wouldn’t be anything but CP/M in this world today,” he later said.

86-DOS was a copy of CP/M-86 to do some degree, but it also included some improvements, most notably in his use an elegant new disk formatting scheme called File Allocation Table (FAT) that still exists in some forms today. It was renamed to PC-DOS by IBM, and to MS-DOS by Microsoft. It would go on to help make Microsoft the most powerful software company on earth, while establishing IBM and its PC architecture as the standard that would drive the next decade of personal computing. It would also trigger the beginning of the end of Digital Research, which would survive but never achieve the same levels of success it had seen in the pre-IBM PC days.

Kildall would later refer to Microsoft’s DOS as “plain and simple theft” in his unfinished memoir, but he had never threatened Microsoft, IBM, or Patterson with legal action, fearing that IBM would destroy his company. Patterson, meanwhile, has always maintained that his work was original, and he defended himself and 86-DOS in an October 1994 letter to MicroDesign Resources after it had published an obituary about Gary Kildall that was cited that memoir and his claims.

“86-DOS is completely different from CP/M inside,” he explained. “It is an entirely original work. Because of the completely different file storage format, none of the internal workings has any corresponding relation to anything within CP/M. I never used CP/M source or disassembly at any time while I was developing 86-DOS. It wouldn’t have made sense to; there was nothing I could learn from it, since my tasks were different.”

Whatever the circumstances, it’s fair to say that personal computing was forever changed by CP/M. All future computers would ship with an OS of some kind, and not just a BASIC in ROM. And the most popular of those in the ensuing decade–MS-DOS–owed its existence to Kildall and CP/M as much as it did to the business acumen of Paul Allen and Bill Gates.