Winner of the 1993 Price Waterhouse Leadership Award for Lifetime Achievement,
Computerworld Smithsonian Awards
Location: Microsoft Corporation, Bellevue, Washington
Interviewer: David Allison (DA)
Division of Computers, Information, & Society
National Museum of American History, Smithsonian Institution
Part 1 Page 2 Page
3 Page 4 Page 5
DA: What would be the two or three things that would characterize the Microsoft way of computer software?
BG: The key for us, number one, has always been hiring very smart people. There is no way of getting around that in terms of IQ, you've got to be very elitist in picking the people who deserve to write software. Ninety-five percent of the people shouldn't write complex software. And using small teams helps a lot.
You've got to give great tools to those small teams. So, pick good people, use small teams, give them excellent tools; vast compilation, debugging, lots of machines, profiling technology, so that they are very productive in terms of what they are doing. Make it very clear what they can do to change the spec. Make them feel like they are very much in control of it.
Have lots of people read the code so that you don't end up with one person who is kind of hiding the fact that they can't solve a problem. Design speed in from the beginning. A lot of things that have helped us, even as the project teams have become larger, and the company has become a lot larger than it was. It is not some methodology where there is a lot of funny documentation. Source code itself is where you should put all your thoughts, not in any other thing. So, our source codes, all though there are a few exceptions, tend to be very well commented in a very structured way.
DA: We talked a lot about some of your earlier successes. What were some of the early things that didn't go well? And what did you learn from them?
BG: There were a lot of missteps in the early days, but because we got in early we got to make more mistakes than other people. I had customers who went bankrupt and didn't pay us. Customers who we spent a lot of time with who never built microcomputer-based machines.
I worked for a long time on an APL Interpreter and I almost got that done, but then it looked like it wasn't going to have much of a market and I was too busy doing other things. We actually never shipped an APL that we had talked to people, that we were working on.
There was another interpretive language called FOCAL that we'd written a version for the 8080 and 6502. Having two interpretive languages like that was not a good approach. That was a dead-end project. Everything else, COBOL, FORTRAN, the way we selected the various chips. A big project we did for Texas Instruments where we wrote a product for their machines which eventually, they failed in the personal computer market, but they shipped a lot of machines in the meantime.
Multiplan, targeting the 8-bit machines instead of just relying on the next generation to come, the IBM PC generation, that was a huge error. When we talk about, "Are we aiming too low, in terms of system requirements, we often think, is this another case like Multiplan?" Because it was a great product, but it was the basic strategy that was wrong. And, in fact, to some degree that allowed me to make one of the best decisions I ever did, which was later, when we had to compete with 1-2-3. There was a question of whether to do it in the character-mode environment, or whether to move up to the next generation, which was graphical. And we said, "Okay, we'll let them dominate the DOS-character world. We are going after Mac and Windows. We are going to be a generation ahead." And that worked out very well. Multiplan was certainly an experience that was helpful there.
In Retail Marketing, we made a number of mistakes that were important for us to learn from. We had in a few countries, agents. And you really don't want to use agents. You want to have your own people. If you are going to be a serious company, take a long-term approach. You should hire people in all the countries you are going to be in and make sure they are there cementing long-term relationships -- not just generating short-term commissions. I think we learned that one pretty quickly.
We did hire in some very sharp business people, and got them to share their experience so it wasn't just us technical guys and the other people. We were very young. I mean, Steve and I were kind of driving the business and Paul and I were driving the technology. We were optimistic in thinking we could get things done sometimes faster than what we did. The project of the moment always seemed very exciting. And some of them never generated much in the way of royalties. But all correctable stuff as long as we sort of wake up and see what the results were.
Relations with Paul Allen and Steve Ballmer
DA: You mentioned your relationships with Steve and with Paul. How did that set of relationships shape and change as the company grew and spread?
BG: Well Paul, of course, was my friend from the early Lakeside days. And we are very close friends today and I'm sure we will always will be [Bill laughs]. He is very idea-oriented. He and I would brainstorm about things. So even though I was running the business, it was a partnership. His role was very, very critical to so many of the transitions that we made. There was always some strain because I was pushing people to work hard, including Paul. That wasn't really a big problem.
When Steve came in, I was spending more time with him because the business side was important; managing and organizing, and (deciding) what we were going to do about international. So, it was great that Steve was smart enough and personal enough, that even though he didn't have a technical background, the programmers accepted him. That was very rare. We didn't really believe non-programmers should manage programmers. And we didn't do that until I think about 1983 before Steve actually directly managed developers. But the developers accepted him early on because he was smart, he would sit and listen to them, understand the things that they really liked to do. And so that fit in. I got a lot of benefit out of Steve going around and always knowing what people were thinking about.
And there were other people. Kazuhiko Nishi, who is a very close friend of mine from Japan, really taught me about the Japanese market. He got us doing the very first Japanese personal computer that any NEC PC-8000 and many of those other projects. He is a visionary, very energetic, almost overly optimistic about where things can go. He started a lot of the early computer magazines in Japan, and worked with us for a long time.
Because it was a fast moving business, although we worked very long hours, we'd go out to movies together. Everybody knew each other awfully well. Up until 1981, Paul and I had owned the whole company, except that we'd shared a little bit of it with Steve. There wasn't much pressure to do anything differently than that, but then around 1981 we decided to share some of the ownership in the form of options. And actually brought some of the really strong contributors into that program.
Vision for Spread of Personal Computers
DA: Bill, you're famous for a vision that you had about personal computers. Can you tell us about the vision?
BG: The vision is really that in the information age that the microprocessor-based machine, the PC, along with great software, can become sort of the ultimate tool dealing with not just text, but numbers and pictures, and eventually, even difficult things like motion video. And that is something that when Paul and I would go around speaking about computers, we would always say that there were no limits. We used to call it the "MiPs to the Moon" speech.
That performance would be unbounded and that all of these incredible things would happen. We were never too specific about exactly when various things would happen. And, of course, when we went back to our business we had to decide what our priorities were. But, the frontiers were sort of wide open. It was that sense of excitement that we really wanted to spark in everybody else wherever we went.
Keeping up with the Competition
DA: Well, you were both a visionary and a pragmatic business person. How did the competition rank in these early days? And how did you manage to continue to make the progress against your competition?
BG: Well, there certainly were a lot of other software companies. Within two or three years of our being started, there were dozens of companies. Some of them tried to do better BASICs. And we made darn sure they never came near to what we had done.
There were competitors in other languages. There was Digital Research, VisiCorp with VisiCalc, MicroPro with WordStar, Ashton-Tate with dBase. And each of those companies made some huge contributions, very innovative things.
Lots of game software companies and that was an area we chose not to get into.
Lots of people doing applications packages like payroll. We'd toyed with getting into that, but decided that it just didn't leverage what we were good at well enough. And that we should probably let other people do those things.
There were many fine companies. I'm probably skipping over hundreds of huge contributors in the software arena. They didn't take quite the same long-term approach that we did. Doing multiple products, really being able to hire people and train them to come in and do great work, taking a worldwide approach, thinking of how the various products could work together.
So, we were more comprehensive. We weren't the largest. There was a time that MicroPro with WordStar was bigger. There was a time when Visicorp was bigger. There was a time when Lotus, with the early years of 1-2-3s incredible success, was bigger than we were.
But we were always the most technical. Whenever anybody else in the software industry wanted to know where we thought things were going, they'd come and talk to us. Because our vision, we shared; we didn't view that as some competitive edge. We just wanted to talk about it and get other people to share the same ideas so that they would help make it all come true.
DA: One of the most interesting machines that came out of this area was the TRS-80 Model 100. Do you want to say a few words about Microsoft's role with that machine?
BG: Yes. This is in a sense my favorite machine, I mean by today's standards it is kind of a pathetic machine. But what happened was Kazuhiko Nishi, my friend from Japan, came over and said that we could have an 8-line LCD with 40 characters. And up to then all we had was four lines by 20 characters. I didn't think using 4 by 20 you could do much that was interesting.
But, when he said we could go 8 by 40, then I got to be pretty fascinated with the idea of a portable machine. It wasn't just taking your desktop machine and trying to shrink it down, because battery life would be a problem, and ease of use would be a problem. But just taking the things you want as you move around and making it pretty inexpensive. So, this machine came out for $500. Jey Suzuki, from Japan, and I, wrote the ROM in this machine. It is a 32K ROM.
Part of my nostalgia about this machine is this was the last machine where I wrote a very high percentage of the code in the product. I did all the design and debugging along with Jey.
And it is a cool user interface, because although most of the code is a BASIC Interpreter, we did this little file system where you never had to think about saving anything. You just had this menu where you pointed to things. It was a great little editor and scheduler. We crammed it all into a 32K ROM. And really designed it in an easy to use way around these special keys up here.
This machine was incredibly popular with journalists. Even though it came out over 11 years ago now, it was out by 1982. You still see some journalists using this, although the technology has gone way beyond it.
We had some great things here like we had a way that you could add a bar code reader to this. We thought maybe people would distribute software on bar codes. In fact, Byte Magazine got into that for a while. We had a lot of ways you could extend this by putting a new ROM in the bottom. And it was sold not only in the U.S. by Radio Shack, but NEC sold it in Japan, and Olivetti sold it in Europe. And the company who made it, Kyrocera, became a good partner of ours for lots of future projects.
DA: You may actually want to turn it on so that we can show it.
BG: Let's make sure that this machine is still running. My God, it's a machine that works! I don't know how LCDs work in a camera. What you had here is just your files. And you would just move the cursor to the one that you wanted and hit the Enter key. And then you'd be back editing that file. So, if we go into text, you can type in the name of the program and it would know that's what you wanted. It is a nice screen editor. You can just move the cursor around.
The only real problem with this product is that the keyboard was noisy enough that if you sat in a meeting with it, it was still considered anti-social because you'd just be tapping away during the meeting. So actually we did a version, just a slight modification soon after it came out, that had a very silent keyboard so that people could sit in meetings and use it. It is really a nice machine. A great, great way that we use these function keys.
End of First Phase of PC History
DA: That kind of wraps up the number of topics we wanted to treat today. Are there any parting comments that you have about this era that you'd like to record before we turn to the PC tomorrow?
BG: No, I think the PC era really can be divided pretty nicely into the pre-IBM PC era and the post-PC era, because as soon as the PC comes out, as nice as all of these machines we've talked about are, they faded from view very quickly. And with the really sole exception of Apple, most of the companies involved faded from view. Most of the software companies who had been focused on these machines, with, of course, the exception of Microsoft.
So, there was kind of a hobbyist focus here, even though we got diskettes, there were some of examples of businesses using these things. They weren't reliable enough or the capacity wasn't enough, or just people hadn't adjusted enough to really think of these as a key business tool. The Apple II is emerging a little bit in that way. But before business would really buy into this thing, it took that next era, which was a whole new level of excitement.
DA: Bill, do you just want to tell the story of this paper tape [referring to the Altair BASIC tape]?
BG: Sure. Besides entering programs into the front panel through the switches, there is always a question of how to send software around. And because Teletypes were reasonably inexpensive, being used in direct communication networks even before PCs came along, the 8-channel paper tape became a key medium of exchange. Now, when a Teletype would read this in, it would only read at ten characters per second. So, when you are loading 8K BASIC it would take about ten minutes for it to read into the computer, and you would hope that it would read in properly.
This tape here is actually the very first BASIC that we did. That BASIC, in total, was about 4K of code. I had to squeeze a little later so that you could run with code and the program in 4K. But at first it was about 4K. So, you see it is about half the size of 8K BASIC. We sold BASIC in this form for most of the first year and a half. This was the dominant form. But then things moved over to cassette and that was the popular form for another year and a half. Then we started to get floppy disk in. We tried a lot of things like reel-to-reel tape, and various digital tape cassettes that people came along with. But it was really the floppy that defined the next level beyond the cassette and paper tape.
DA: You were telling us yesterday that you actually stayed up just about all night working on this tape. Do you remember what it was like to actually punch that tape before you sent it to Albuquerque?
BG: Well, the PDP-10 that we did the development on actually had a high-speed tape punch, so it could punch the tape in about three minutes. So, the issue wasn't so much the tape as it was making sure everything that was on there was right. I actually used fan fold paper tape, it is the same, but they fold it every six inches or so. And I punched three different copies out on the PDP, which is the peripheral name of a paper tape punch on the DEC 10. I made sure that Paul had all of those in case there was any kind of problem. But paper tape was actually quite reliable. It was just slow to work with and easy to mishandle. I mean you have a huge paper tape. You're always kind of dropping it. And it would get twisted up, things like that.
Holding the Beginning of Microsoft in his Hands
DA: So, really, what you have in your hands is the beginning of your company.
BG: Yes, this BASIC was the first real piece of software ever written for a PC. And it became for the first generation of PCs the thing that unlocked the power that was there, because, although some people did machine language programming, 90% of what was done was done in BASIC. And 90% of that was Microsoft BASIC. But, descendants of this tape got onto all those early machines.
DA: Bill, when we left off yesterday, we were talking about the PC, the IBM PC, as beginning a new era in the history of the personal computer. Why is that so?
BG: Microsoft started focusing on developing 16-bit software as early as 1979. We decided that what Intel was doing with their 8086 was really the way to go. Actually, Motorola had played around with the 68000 design and decided not to go forward aggressively with that. Instead, they did the best 8-bit processor ever done, the 6809. And we did a BASIC for that. I liked that chip. But in terms of really going after a larger address base, this 8086 family was it. And so even before IBM came to us, we worked with a number of companies, including a little local company called Seattle Computer Products or SCP, to show that we were moving software over and writing new software for the 8086. It was higher speed and it broke the 64K barrier of the previous machines.
We also thought that in moving up to this new level, instead of having each manufacturer go and do different features so the software's special, we'd have a chance to set a binary standard so that the same system software and all the same programs would run on one machine. So, we were looking for a chance to do that.
The first companies to do 8086-based machines, big companies, were Mitsubishi in Japan and Victor. Both did very nice 8086-based machines with good graphics. But the seminal event certainly was the group at IBM that pulled the machine together.
It is kind of a long story that has been talked about where they were chartered to do a home PC, but even more importantly they were chartered to do something quickly. IBM was frustrated that they were taking five years between product concept and shipment. So, this group had sort of bid to get the charter to do the PC based on saying that they could do it in less than two years. In fact, they had said they could probably do it in a year. But the trick was to go to outside suppliers for the key components.
So, they went to Intel to get the chip. And they came to Microsoft to talk about, not only software, but the overall system design. They'd seen that we were in all the personal computers up to that point, and read about us before they came up. Told us that we had to sign quite a non-disclosure. These were just product planning people. But we talked about our enthusiasm for a 16-bit PC. That they could go and told them they ought to talk to Intel and it could be even better than the Apple II or the CP/M-80 machines which were sort of the higher-end popular designs of that time. So, the project started in late 1980.
The very first issue of PC Magazine came out with the story about the work between IBM and Microsoft. Dave Bunnell had been at MITS on the Altair and I had gotten to be good friends with him. He actually started up a PC magazine after he left MITS, Personal Computing, and I had done a column for him. So, when we were working on the development of the PC, I encouraged him to go out and do a magazine around it, and he worked, actually, with Tony Gold who'd had LifeBoat Software, and started up this magazine. So, he came out and interviewed me and I talked about the whole project.
It really started when they came out in the summer of 1980. It is pretty amazing. They came out in the summer of 1980. We talked, agreed to work together by September. Started engineering work. Actually signed the contract in December, froze the ROM, which was the BASIC built-in to this, there's a ROM BASIC in here [gestures to the IBM PC] in March and the product shipped into the market in either August or September of 1981.
They had actually a fairly small team. So, in a sense we were working with IBM, but in a sense working with a small team. The small team included people who had used PCs before. The Chief Engineer of this, and the person who deserves, at the hardware level, an immense amount of credit is a guy named Lew Eggebrecht. He later got into some disagreements with IBM and left and sort of never got his due, in my view, for his role.
The team, of course, was headed after Bill Lowe got the charter. It was handed over to Don Estridge. And so it was Don who I worked with a lot, and the people who were working for Don. It was a very close partnership at that level because they were a little naive and very open-minded. They told us for this keyboard, they said Lexington had to use a certain layout. We wanted these function keys to be up here, that the Lexington layout forced them to be over here. We ended up with this funny big key over here because of Lexington. But here we'd put in our favorite fully-extended character set. Came up with some new ideas there. Put some special things in here that we thought we'd be able to use. In fact, we put in the Wang word processing character set because we weren't sure whether to do a Wang-clone word processor, or start with our own approach at that time.
IBM-PC Design and Development Issues
There were many issues about this machine. Should it have graphics or not? First design did not. We convinced Eggebrecht that he could do a graphics card. The funny thing was, they wanted to have these good looking characters. So, they did a video card, this character mode card that didn't support graphics, but they did go ahead and do the color graphics adapter, CGA card. The graphics weren't very good, 640 by 200. The characters were only 8 by 8. So they don't look as good as this character display does. But that was fundamental. If they hadn't done that, there are so many of the great things about the PC that wouldn't have been possible.
DA: What was it like to work with IBM? They were a very different company from the companies you had worked with before.
BG: Well, IBM, as a big company, was very different. It meant the non-disclosure, the contract, the security measures, the way we related to their various overseas groups. That brought in "the Big IBM". But the individual people, whether it was Sandy Meade, Mel Hallerman, Dave Bradley, they were all reasonably similar to us. A little bit older, but also pretty hands-on, pretty energetic. And a small group did this.
We actually had more people assigned to the project than IBM did. And we were only a company of 30 people. We threw ourselves into this. Although we had first just started talking about the BASIC, but then they said they wanted other languages. We knew that we had helped design a machine that was more than a home computer. So, then we were kind of saying, "Well, if you want the other languages, you must know too, this is not just a home computer. Let's not joke around." You don't put COBOL and FORTRAN and all that on this machine.
And so, that group and Microsoft sort of conspired to make it an all-purpose machine, not just a home machine.
The base machine came without a floppy disk. No disk at all. And 48K of RAM and the ROM BASIC. They didn't ever sell many machines, but actually in terms of designing this thing, that was very tricky to get everything to work in that much memory and have the ROM so it could work without the operating system there, but also work when the operating system came in and yet be patchable to the ROM if anything was ever wrong. Very tricky issues.
Some interesting things include: the way we had decided that the schedule was so tight that we would refer IBM to go down to talk to Digital Research about CP/M. Well, they didn't want to sign the non-disclosure and they really didn't jump on it. So, the project was at risk. It kept having to go through IBM reviews. And if they didn't have all of the software signed up they clearly weren't going to make the schedule.
So, we said to IBM that we could do that. And as an increment on top of what we had committed to do, it was about ten percent extra work. So, we went out and bought from Seattle Computer Products the work that they had done on an operating system they called the SCP-DOS or 8- DOS at the time. But, more importantly, we got Tim Paterson to come across and work for us. He was the primary creator of MS-DOS. Bob O'Rear was one of our people who worked very closely with him and put that together.
I was very involved in the creation of the BASIC. Actually, Paul Allen got involved in that, and a number of other people. But, when it came right down to it, everybody pitched in, Neil Konzen, myself, Paul. There was a very tricky design. For example, proving to them that we could do some great stuff in the graphics area so that it was worth doing this CGA card. Proving to them that we could use the palette. We were able to take a lot of innovative work we'd been doing on machines in Japan, OKI F-800, Hitachi BASIC Master Level III, NEC PC8800 and take that and put it all together here because this was a 16-bit machine. And so it came out very, very well.
They didn't invite us to the introduction. That was kind of a unique IBM thing. But we had a great relationship. They were down in Boca Raton. So, we were taking lots of all night flights down there. They kept the machine locked away in this hot room, which was a real problem in terms of working with it. Whenever there seem to be security leaks, they'd come and ask us. Actually, Kazuhiko Nishi did some of the design. He made sure they put speakers in. We had a primitive sound capability that was Kazuhiko's idea. So, they were worried -- were the leaks in Japan due to him? Very intense project.
DA: We should hone in a little bit, Bill, on the MS-DOS development. You wanted to show us a little bit of that. Your involvement with MS-DOS shaped a lot of what would happen later. So, let's just talk a little bit about -- you've highlighted the issues going into it. But just taking some of the fundamental ideas were that reflected MS-DOS that are still part of personal computing today.
BG: MS-DOS, although in its first version was fairly limited. It had to run, as I said, in that 48K machine. It went on to become the foundation for this binary standard where so many applications would come along. It was not the only operating system brought out with the machine. They actually brought out CP/M-86 which was when Digital Research woke up to 16-bit, what they went and did. And another thing called the UCSDP system.
But the one that shipped first with the machine and the one that was low priced, because we didn't insist on significant payments, because we wanted it to get out there, was called the IBM Disk Operating System. Now, the funny thing is IBM didn't like these acronyms. So, in fact, even the term PC or words like PC DOS, they at first really didn't like people using those. But it became so commonplace that the operating system was sometimes called PC DOS, sometimes called MS-DOS. And, of course, the machine became the PC and all the magazines about it were the PC magazines.
We worked with IBM to have some software available from the very beginning. We are seeing here some of these BASIC sample programs. Neil Konzen and I spent a whole weekend just goofing around with this machine once we had BASIC running and wrote this thing, DONKEY.BASE, PIECHART.BASE, these are still shipped today. They are kind of obsolete, but they are a code that we threw together that weekend.
The key to the competition in operating systems was getting lots of applications. And Microsoft got very serious about working with manufacturers and software developers. MS-DOS started out a little bit behind because it was easier to take 8-bit programs from CP/M-80 to CP/M-86. So, we came up with tools to help with that.
And it was a real competition, I mean although it is an obscure footnote in history now, from 1981 to 1983, people weren't sure which system would win. Even until as late as 1986 some companies like Digital Equipment with their machine, the Rainbow, were very focused on CP/M-86 and not on Microsoft MS-DOS. And then by 1986, Digital Research kind of stopped really trying to compete and MS-DOS really pushed ahead.
Part 1 Page 2 Page 3 Page 4 Page 5
David K. Allison, Chairman, Division of Information Technology and Society. National
Museum of American History.
Research specialties: Computer technologies; military technology; social history
of technology.
B.A. (1973) St. John's College; Ph.D. (1980) Princeton University.
09/27/1994 Final Edit Kris Kaeding
06/28/1995 Webification by David Allison and Mikel Maron
Copyright 1994