| Transcript of an Oral History with Ben Slivka
Interviewer: Lee Dirks, Archivist, Information Services, Microsoft Location: Microsoft Studios, Microsoft, Redmond, WA Date: November 4, 1998
|
LEE DIRKS: Today is Wednesday, November 4th, 1998, and we are at the Microsoft Studios on Microsoft Corporate Campus in Redmond, Washington, and interviewing Ben Slivka for an oral history with the Archives.
Ben joined Microsoft in 1985 as a software design engineer, and worked on numerous products and projects, including MS-DOS, OS/2, Windows 95 and the Microsoft Java Virtual Machine. Perhaps his most high profile contribution to the company has been his breakthrough work on the Internet Explorer, where he led the creation of this product in version 1.0 up to the very successful release of IE 3.0 in 1996. Since this time, Ben led Microsoft’s Java efforts for a year and is now working on designing and prototyping a new Windows user interface for future versions of the product.
LEE DIRKS: So, Ben, could you start off with giving us a little biographical background on yourself? Give us details about where you were born, and growing up?
BEN SLIVKA: Lee, it’s great to be here. Yes, I was actually, coincidentally born in Seattle, 1960, at Swedish Hospital -- I guess where Bill was born, too. Grew up through high school, went to Seward, Madrona and Meany middle schools and Garfield High School.
I went off to Northwestern in 1978 in Evanston, near Chicago. That was the first airplane flight I’d taken and the first time I’d been east of the Cascades. I got my undergrad degree in applied math computer science, and then went to work for IBM for some reason at Poughkeepsie, New York. They were doing a supercomputer kind of vector machine, in ’82. I got there and it was like, wow, who’s idea was this? Actually, I had gotten married my senior year in college to Lisa Wistner(?), and she transferred to Vassar, because she was…well, I was a senior, she was a freshman. That’s another story. So she was at Vassar, I was at IBM, we were both there in Poughkeepsie, and I’m going, what the heck, who’s idea was this? So we barely lasted the year.
We trenched back to Northwestern, and I got a job full time at the computing center. She finished up her undergrad, and I got a Master’s in computer science, because it was like a hundred dollars a course. So, having had the IBM experience, where there were very few smart people there, and I thought, okay, mainframes are dead, minicomputers are next--okay, go do the microcomputer thing. So in ’85, actually in ’84, I sent a letter to Microsoft saying I am great, hire me. I got a letter back from Joanne Mahall(?), who was the recruiter at that time, saying send us some source code sample. So I sent off several bits of source code. Then they said, okay, fly out. So I remember getting out…actually, stayed at my parents’ house. They lived in Mt. Baker, and I flew in like the last day of January. So it was like February 1st, and I wake up and there’s snow all over the ground. I’ve got a rental car, front-wheel drive, little Japanese something.
So I’m cruising around the streets at 35 m.p.h., because I’m used to driving in New York and Chicago, and all the natives are just strewn, and I’m just passing everybody. So I get to Microsoft February 1st, and there’s all these 10-year-anniversary things around. Of course, there are not very many people on campus, because of the snow. This was back at the Northup buildings.
So anyway, I did my interviews and I remember that Gordon Letwin was the…as appropriate interviewer. I’d sort of seen some stuff about him, so he had this little crystal ball thing. So we sat down and he asked me some questions about stuff, and I remember he had also programmed on Control Data Corporation mainframes when he was at Purdue, I guess. I’m out at Northwestern. So he asked me how would you implement a stack on that machine, which didn’t have a stack. So I don’t think I answered that very well. But, anyway, I guess I did well enough for the day that I got to Joanne’s office and she gave me the offer. There was stock options, and she was like, we have these…they might be worth $10,000 or $20,000 or something. I’m like, I don’t know what these are, whatever. So that’s kind of how I came to Microsoft. That’s kind of broad background stuff.
LEE DIRKS: When had you first heard of Microsoft?
BEN SLIVKA: Hmm, that’s a good question. Certainly by, let’s see, the Macintosh came out in ’84, and Northwestern was one of these Apple University Consortium schools that got a price break on Macs. It was Apple’s marketing strategy to seed the universities. So I’d certainly heard about Microsoft by that time. I’m not sure if I really knew about Microsoft before that.
It’s sort of funny, because when I was growing up in Seattle, I got into computers in like ’76. Of course, back then, getting into computers was a hard thing to do. So HP had these little hand-held programmable calculators that had sort of 49-step instruction memory. I used to go down to take the bus--I didn’t have a car--go down to the old Frederick & Nelson building in downtown Seattle, which is now Nordstrom’s flagship building, and there was a little calculator department. So I’d sit there and write little programs on the calculators, and the sales ladies would let me do that, as long as there weren’t customers around.
I also used to go…and I lived in Mt. Baker…so I used to go out…there was a store called the Retail Computer Store near Greenlake. So it was an hour and a half to two hour bus trip with all the transfers, to get out there. I used to go out there. This was back when the Altair and the IMSAI and the SOL-20 and all those first microcomputers were available. So I’d, with my scarce savings, I bought a floppy disk, and would like go type in basic game programs, or something into the SOL-20 and save them carefully on the little disk and take it around. Again, the salespeople there would sort of humor me, as long as there weren’t any paying customers around. I never figured out that maybe I should have gotten a job there. I never quite did that thing.
Like my junior or senior year, I got an account at the University of Washington. They had an academic computing center, which had this Control Data Corporation hardware, which was my first experience with that. So I’d write Fortran programs and write Pascal programs. Actually, my high school had…I think my junior year there was a desktop programmable calculator, which had like a paper tape, not like an adding machine tape, was its output. Not LEDs or CRTs or anything--it was an adding machine. It had basically a very simplistic keyboard. But I could write little programs. I wrote like a biorhythm charting thing. There was a plotter attachment, and I wrote a little race program, which you could draw a race outline and move the plotter pens around. There was actually an assembly language for this thing which you could get access to. At one point, I remember I literally dumped the ROM out with ka-chunk, ka-chunk, ka-chunk. It puts these things out. I punched cards at the University of Washington. So I literally took a ROM dump of the whole thing and then wrote a dis-assembler, so I could spew out what the ROM was and try to figure out what was going on there. I never got all the way done with that.
So anyway, that was some of my early computing, and I did some, actually did some I guess what you might call consulting. There was a Ph.D. student who needed some work done in statistics, so I learned how to use SPSS and did some statistics stuff for him. My dad was in the Seattle Symphony, and in the Musicians’ Union, and they were trying to negotiate a contract with the Symphony. So they wanted to do some projections forward about what kind of funding they needed for their pension plan. So something today, which you would have whipped out Excel and done in an hour, I…of course, I had to punch up all the cards with all the information about each musician. Then I had to write…fortunately, I’d programmed it to basically do the forward projections and do all the compound math, assuming stuff. So that sort of funded my computing habit, which was…I spent a couple of thousand dollars a year or something in computer time. Today, of course, you could buy, I don’t know, a Pentium 300 and compute ‘till you’re blue in the face.
First Job at Microsoft -- Building DOS 4.0
LEE DIRKS: So, you described your first day at Microsoft. What was your initial role, when you started your work? What position were you hired for?
BEN SLIVKA: I was hired into the DOS 4 team. I started June 24th’85. This project had been underway for several months, and the idea was to build a multi-tasking operating system. Still it ran in 640 K of RAM on basically an IBM PC, so not even an XT necessarily, 8088 running at 4.77 megahertz. So we were making progress on that, and then we were trying to get a deal with IBM. So this thing eventually became OS/2. So we, at some point, realized that the 8088 and 640K of memory wasn’t going to be enough, so we moved to the 286 and sort of started all that work. I worked on OS/2 ‘until September of 1990.
Most of the work I did was actually on the DOS compatibility stuff. It was kind of a challenge, because the first several versions of OS/2 were targeted at the 286. So there was this idea…so you had to be able to…you wanted to be able to run protected mode programs, so they could get access to more memory. Yet you also wanted to be able to run DOS programs that had to run in the real mode of the 286, and it didn’t have this nice virtual 8088 mode that the 386 and later processors had. So there were these goofy architectural decisions we made. We had this idea of dual mode device drivers. So things like the disk driver, the keyboard driver, the mouse driver weren’t a standard thing at that point, but eventually were. All these device drivers that took interrupts and had to respond to devices, they had to be able to execute both in real mode, where you take the segment in offset and the segment ships over four and gets added with the offset, and the protected mode, where the segments are just selected to some arbitrary place in memory. Anyway, so that was kind of a challenge to be able to write code like that.
There were lots of performance issues about switching between real mode and protected mode. The 386 wasn’t designed to go back to real mode, so Gordon Letwin came up with this way where you basically faulted the processor and caused it to reset itself. Then there’s a way to hook into the Bios to get control again. So we did this sort of unnatural thing. He got a patent for it, actually. Well, we can talk probably more about OS/2. So I did DOS compatibility stuff. I worked on the file system, some performance things. Then for OS/2 2.0, which was the 386 version of OS/2, led again the DOS compatibility stuff, then sort of switched half way through and worked on the visual shell and printing stuff.
BEN SLIVKA: Oh, let’s talk about OS/2, though, because I think we have some things we can learn from that. The one thing was, the way I like to describe the OS/2 project is two guys get in a car in Los Angeles. They start driving toward the east coast. They want to make it there pretty quickly, so they take turns driving. So they progress that way across the country. The problem comes up that one of them is trying to go to Florida and the other is trying to go to New York City. So the real problem with OS/2 is that we had this desire to get productivity software like word processing and spreadsheet out there. We wanted to do graphic user interface. Those were our primary focus. We wanted to make it easy to use and powerful, and all that.
IBM was a lot more concerned about protecting their mainframe business, so they wanted this OS/2 to be a great terminal to connect to mainframes. So we just never had any way to resolve those fundamental tensions about the project. So in September of ’90, we finally said we’re going our separate ways with IBM. We’re going to just focus on this Windows thing.
But I remember several years before that seeing Ballmer in the hall and going, Steve, just divorce those guys and let’s go do this ourselves. So OS/2 was a pretty hard thing. It was fortunate that we did have the Windows group doing that other stuff. I remember David Weise, who was in the Windows group -- we'd bought this company Dynamical Systems Research that Dave Weise and Cameron Mhyrvold and Nathan Myhrvold and a couple of other folks…there were just five of them -- he was always, Ben, come work on Windows. I’m like, oh, well, sort of a loyalty thing. Or, no, we have to kind of get this OS/2 thing done. But, in retrospect, it would have been much more fun to be able to work on Windows. So I think the main lesson to learn from the OS/2 thing is, when the goal isn’t really clear, it’s very hard to be successful. The project was just compromised on a variety of fronts. IBM didn’t want to move quickly to do our 386-specific version. Then they had these goals about mainframes and connectivity, where we were focused on personal productivity and GUI.
LEE DIRKS: So, in a certain way, Microsoft gained its freedom. But in another way, we lost a big contract. What was the feeling around Microsoft during that time, when we announced the divorce.
BEN SLIVKA: Well, I remember there was about a month where…actually, so I was running the visual shell team. We’d done this work to do this new shell for OS/2.0. Many of the ideas actually were shipped in Windows 95. I remember we did this demo for Bill around July of 1990, and Windows 3.0 had shipped at this point. I just remember kind of the sense coming out of the meeting was, well, that’s really nice, but it’s not Windows. So already at that point there were these little hints that Windows was going to be the thing we were going to be doing, going forward. So…and there was the last month, like that month of September, people sort of knew that like this OS/2 thing…it was just a matter of when was the shoe going to drop. So people were actually off writing Windows programs. I told everybody, hey, you know, go write some Windows programs and learn about that. I think, so, really, I think the OS/ 2 team mostly believed that, you know, OS/2 wasn’t successful in the marketplace. We were getting a lot of negative feedback from press and analysts and reviewers. So I think we were relieved that we got to give up on that thing and go focus on Windows. So I think people were mostly excited about that.
So the next thing I did was this WIN 32 for DOS project. We decided, okay, we’re not going to do the OS/2…so Windows has, you know, users…things like dialogue boxes and menus and windowing and all that, and GDIs graphics device interfaces, drawing text and lines and colors and things on the screen. The OS/2 had this presentation manager set of API to do similar things, but they were similar but very incompatible. The other thing happening in 1990 was that we started this NT team, back in I think ’89 or maybe as early as ’88 with Dave Cutler and some of the people from DEC West--Digital Equipment Corporation. We had hired them to go build this new technology operating system. So they’d been going off, and this was called OS/2 3.0 at the time. Then we said, okay, we need to find what the programming APIs are for applications. So I had about 12 people working for me, and then there were people on the NT team. Together, we basically put together this WIN 32 stack. So what we basically did was take the WIN 16, the API spec for user, GDI and kernel, and sort of widened all the 16-bit parameters to 32 bits. There were some things where we had to kind of shuffle things around, because there had been various parameters that were packed, you know, two 8-bit parameters packed into 16 bits. So we had to kind of change some things there. We tried to define those API in a way that it was very easy to port a WIN 16 program to the 32 bits.
We spent three or four months kind of doing those specs and doing those. Basically, the NT team were sort of coming from more of a minicomputer focus, where a lot more memory is available and that kind of stuff. We were bringing the sort of small memory size perspective. So we got those things defined. But, otherwise, that WIN 32 DOS project, we weren’t very successful, because we didn’t know what we were building or when we were building it or who we were building it for. But, other than that, it was a great project. So we defined these WIN 32 APIs, so that was a good thing. That kind of meandered around, and eventually, when WIN 3.1 shipped, you know the WIN 3.1 team took charge and did Windows 95. So they were able to leverage the fact that we designed these APIs, but we didn’t really leave them much code that was of much benefit.
Now, we were up to March of 1992. At this point, okay, what’s my mental state? I’ve been at Microsoft almost seven years and I haven’t shipped any product that’s been successful, right? I’ve gotten these stock options and that’s a good thing. But I came to Microsoft because I wanted to ship stuff. I just liked to build stuff. When I was small I built lots of stuff out of Legos and I built model rockets. Then I found computing where you can build things, unlike model rockets where if you get the fin glued on wrong, you have to scrape it off, and it’s never going to be as good. Painting gets messy, the fumes. With software, you can build anything and it’s less messy. In March of ’92, I’m feeling unfulfilled. I haven’t shipped anything that’s been successful, yet I’ve gotten all this compensation, which is a good thing.
So I just want to ship something, anything, I don’t care what. How about MS DOS 6? Sure, that’d be a great thing to do. You have to understand that at this point, in ’92, people are focused on Windows programming, there’s this WIN 32 thing coming up, people are working on Windows 95. So here I’m going to go do MS DOS version 6, which is character mode, it’s 16-bit. Like we’re using assemblers and compilers that nobody even works on anymore. But it was a great bit of fun. We basically looked at…so I was basically the first developer on that.
We looked at what were the kind of the some of the very popular utility programs that people were using with MS DOS. Those were things like 386 memory managers to sort of move device drivers out of the 640 K space. There were disk utilities for doing defragmenting, for checking the status of the disk. There was this product from Stac Electronics, called Stacker, which was a disk doubler that did sort of on the fly data compression. Things like backup, and there was antivirus--a whole bunch of utilities. So we basically picked a set of those things to go integrate with the operating system and provide a lot more value to customers. So MS DOS 6 was a very fun project, and I spent a lot of time on the disk compression feature, which we called Double Space. So we evaluated a couple of companies for their technologies, and then decided to license some technology from this company, VertiSoft, that had a product that I think was called Double Disk.
So then we worked really hard to, you know, how was the user interface going to be? How were we going to make it simpler? How was it going to be better than standalone products like Stacker? We started in March ’92, and we got our first beta release out I think in October with disk compression working. Then we shipped at the end of March of ’93. So that was a pretty fun project.
One of the things we focused on was could this thing be very safe. So, for example, with the Stac product, if you were compressing your drive installing it the first time, if the power went out, your system was basically unusable. Hopefully, you backed up your data, because you were just left in the middle. There was no way to pick up the pieces. So a thing we did, we said that scenario…if the power goes out, when you turn your machine on, will pick up and keep going. So, in fact, we had that as a design principle, and we made sure that basically at every point during the compression we knew where all the data was. If a machine got recycled, power cycled, you could get started again.
Another thing we did was a lot of testing. We would…we wrote a little utility that took an image of the entire disk and wrote it out to the network drive, so that if some problem happened, you could just spit that back onto the machine as if nothing had happened. Starting at about seven at night, after people had gone home, we’d go pick various groups like the Exchange group or the Legal group or the Business group, whatever. We would go out and take an image of the disk, and then we would back it up, using our backup software that we were shipping. That was a good way to test the backup software. Then we’d do this compression, and sometimes we’d flick the power on and off a couple of times just for fun. Then we’d do what’s called a backup compare against that backup we’d made--not the disk image, but the backup, just to make sure all the files had gotten transferred. Anyway, we did that for, I don’t know, 700 or 800 or 900 different machines. That’s a pretty laborious thing, but the other thing we did is we kept all those disk images. Then we had an automatic process, when we had a new build out, we’d have these test machines set up. They’d go grab an image, slam it down, do those steps, make sure everything was okay, and then get the next image down. So we had a nice kind of automated way to make sure that we weren’t going to lose any data.
So that was kind of a fun thing. It’s funny, we got a piece of e-mail from one of our beta testers, you know. Yeah, I backed up my system and I started compressing, and then I went off to like change the light bulbs or light fixture in my room. I flicked off power things in my circuit box, ooh, no, I realized I turned off my PC. I turned it back on, and went and got my backup tape and I went back to the PC, and oh, it’s still going. What’s going on here? So I let it finish, and then I did the compare--backup compare. Wow, it did it all, amazing. So it was sort of this thing we designed. Someone actually had it happen to them. So that was kind of a fun thing.
Of course, there was a negative thing that happened with MS DOS 6, which was we got sued for patent infringement. So this Stac Electronics…so that’s a really long story. I’ll give you kind of the summary view. Basically, they had one patent and they sued one company in the summer of ’92, we released in March of ’93. We looked at that and we’re like, okay, we’re okay with that patent. We made sure we weren’t going to have any problems with that patent. Then they sue us on January 25th of 1993. So it turns out, unbeknownst to us, there was a patent referred to on their original patent that they bought in November of ’92 for $200,000. That was a much more interesting patent. When they sued us, we saw that patent, and we engineered around that. There was a big rush effort where from the end of January to the end of March, those two months, I didn’t get much sleep at all. There were people from the Research group, Rick Rashid and John Miller, and then some people from the Windows team, Jeff Parsons, and also Mark Zbikowski. So we had this thing where I would like code up sort of algorithms and see to make sure that they had the right kind of compression properties. Then Jeff Parsons or Mark Zbikowski would go implement them, finely tune in X-86 assembler, to make sure they’d be fast enough. So we did all that kind of stuff, so there were five ways we were different from five of the claims that they had in this other patent. So then, the trial was actually the following year. So we actually changed our stuff and shipped it. Then the trial was early ’94. In fact, the week after my second son was born, I had to go down to Los Angeles for two weeks. Like what could you do?
So we had the trial. The jury was picked the day after the L.A. earthquake in downtown L.A. Okay, so you have to ask yourself, who is going to go to downtown L.A., where there are some tall buildings, like the day after an earthquake, you know, kind of a major earthquake. That was the jury pool. So the trial went on, and we testified on a variety of things. We counter-sued them for our patent thing and that they’d stolen our trade secrets about integrating data compression into our operating system. So the verdict came down, they were guilty of stealing our trade secrets, but we were guilty of infringing their patent. Then there’s a court newspaper scandal sheet or something that goes around. They’re allowed to ask the jurors what went on. Actually, the attorneys from our side and from Stac’s side aren’t allowed to talk to the jurors. The jurors came down to like, oh, you know, it’s like it was David and Goliath, just the way the Stac attorneys had spun the whole thing. So they didn’t…like, did they understand what hashing was, or patent claims, or…they didn’t understand any of that. They were just like, well, it seems like the big guy whacked the little guy. So we felt confident we would’ve won on appeal, but that was just going to take too long. The judge is like doing enjoining things. So, anyway, so that whole thing was kind of a mess.
I don’t know, what did we learn from that thing? Pay really close attention to the patent stuff. You know, the hard part was it was hard to know, because we’d designed around these patents, and we felt confident that an appeals court that actually understood patent law would’ve seen that. But it was too subtle for a jury. So, anyway, yeah, jury trials are pretty interesting.
Then another bad thing happened was that InfoWorld published an article saying MS DOS 6 causes data loss, and people were concerned. Like why should I compress my drive--what if something’s messed up? I won’t be able to get my data back. So they did this test. They actually called us up on a Thursday…or, they called us up on a Friday and said, hey, can you comment on this story? We’re about to run this story that MS DOS 6 causes data loss. We said, well, really, well, can you tell us more about…you know, can you hold off and let us come down and see what’s going on? No, no, no, we’re going to publish the thing.
All right, so we went down on a Saturday anyway. We found that they had this network testing lab and they had really two problems that they encountered. One was that with MS DOS 6 Smart Drive, which did this sort of right behind disk caching, was installed by default. And that had been included in Windows 3…let’s see Windows 3.1…no, Windows 3, in 1991. What they were doing, they had…in their testing program, they had a utility that hard-rebooted the machine. It didn’t go through the standard software reboot hooks. So Smart Drive has a way to catch that and flush its cache before the reboot happens. But they were skipping out of that. That’s not something a user would do, right? And that’s not something a corporate customer would have this program that hard-rebooted the machine. This was some utility they’d written that didn’t really conform to the standards of PCs. So that was their fault. They were causing data to get thrown on the floor, and that wasn’t a customer scenario.
The other problem that they had was Lotus 1-2-3 for Windows was getting some fault. That turned out to be again related to Smart Drive. Smart Drive took up more memory, and there was a bug in Lotus 1-2-3 for Windows in low-memory situations, where it would have these errors. So none of these things had anything to do with data loss or anything that customers would experienced. Anyway, but they published this story and they had all these machines with red Xs, because they’d run the same two tests on like 25 machines, or whatever. Then like every week after that, they ran this story, MS DOS…you know, they just kind of got on this little jihad about that. So then we turned around and did an MS DOS 6.2, and…so, there were no bugs to fixed. So we looked at all this, where’s the bug? There’s no bug to fix.
But we did things like we added, well, this scan disk program, which was like Norton Disk Doctor, a very exhaustive disk scanning thing, which was a good program to write, and it was a good thing we had it. We actually ran that before we would compress your drive, because we found that some people did have like…if you had a bad hard disk control, you know what? Surprise, if you try to compress your drive, it won’t be a happy thing, or, if you have a bad disk, or things like that. We found there was one TSR program that, if you did a write of one byte to a file, it would actually cause the file to get truncated at that point, just because of the way MS-DOS works. They hadn’t really thought that through. It was a TSR program. So that would cause us to truncate our compress volume file. That was a…how could we know about that, that problem. So, we changed it to two-byte writes. Every write was two bytes. There were some other weird kind of things.
We wrote a…we found one problem with one particular machine where it had a bug in its memory controller. So we actually wrote our own…this is how paranoid we were, you have to understand. We wrote a memory test in the high mem.sys file that basically filled all of memory with this very distinguished pattern. Then we went back through and read it again, to make sure nothing had changed. Because, what had happened in this particular machine is, you write one place and some bits change someplace else. So we put that program in. We wrote…we had this other feature we called Double Guard in the Double Space driver, where, to be faster, it reads some of these data compression structures in the memory that keep track of where stuff is, so it doesn’t constantly have to be reading those off the disk. So there’s sort of a cache.
We were hypothesizing that maybe those get corrupted. So, we wrote this thing that brings them in, check sums them, and then, before we read them again or modify them, we’d do the check sums again and make sure something didn’t get trashed. We actually found bugs in a couple of network drivers where they were trashing memory. So that was kind of a paranoia we had about that. But I bring that story up because it was really thinking about the quality and making sure that the system worked.
LEE DIRKS: So you shipped 6.0, you shipped 6.2, and then what was your next position after that?
Data Compression -- Codenamed Diamond
BEN SLIVKA: Then I did a little, I looked around at a couple of things. I did an independent study project, which was this…Diamond was the code name for the project. Basically it was, the way we’d distributed software before that, we had a data compression program that compressed each file individually. The compression wasn’t as good as PK Zip, which was the benchmark at the time. Then each group, whether it was Excel or Word or the Office team or MS DOS, had their own little custom ways, spreadsheet programs or VB programs to bin pack the files on the floppy disks.
So I wrote this thing that you basically give the whole list of files, then compress them all together as a stream so you could get compression across multiple files, then chopped it up into sort of floppy disk-size units, or whatever size you specify. It was all configurable. Then I licensed a compression algorithm, which was better than PK Zip by 20 percent. Then a third thing, I hired a guy from Iomega, and he came up with this 1.68 meg floppy format. All those things together saved us 20 or 30 percent on the number of floppy disks that we had to ship. So the first couple of years, in ’94, in ’95, in ’96, until we sort of shifted to more CD ROMs. We probably saved, I don’t know, $60 to $90 million dollars a year. So that was kind of a fun independent study project. That was just a thing that I just…you know, I thought it might be a good thing to pursue, and I convinced I think Silverberg or Paul Maritz that I could go do that. We just went and did that.
BEN SLIVKA: Oh, yes, the Internet, okay, good. This Diamond project’s winding down. It’s summer of ’94, and Brad Silverberg does a very smart thing. He pulls several people off of the Windows 95 project, in the summer of ’94, to go think about what to do after Windows 95. We’d never really done that in the Windows group before, to have a group of people plan the next version before the current version was done. There were several people--John Ludwig, Thomas Reardon, myself, Rob Price, maybe a couple of other people. So we tried to go like what are the investments we should make in the next version of Windows? So we looked at things like, obviously, the Internet was one. We looked at computing in the home. We looked at game machines. We looked at wireless networking. That was some idea that everything would be connected in a wireless way. We looked at the problem of how could you get the Win 95 and the NT code bases together, because there was this desire to have one desktop operating system. Those were certainly the main topics. We might have looked at a few other things.
So, out of that, so I got kind of excited about the Internet. I remember, I think it was…I saw a browser, like Phil Barrett, who had managed development for some of the earlier Windows releases, was I think working on a Web browser at this time. I saw Mosaic or something in his office. I went, wow, that’s pretty cool. Today, of course, we have proxy servers and you can gateway out to the Internet from your desk in your office here at Microsoft, but at that time, there was no way to do that. I didn’t think about getting like a phone line and a modem and ISP--I didn’t think about that. But we did have…there was a separate network inside Microsoft to get out to the raw Internet. So I got a "direct Internet tap" installed in my office. I had to get through some hoops to do that. I remember in late July of ’94, I got an Internet tap in my office, so I could surf the Web from my office. I had a separate machine hooked up to that.
I thought, hey, that’s pretty cool, we should go build one of these things. So I’m kind of finishing off the Diamond stuff and then Phil Barrett is not making very much progress. I’m helping him. I’m categorizing UI stuff and looking at what Mosaic and some of the other competition are doing. Finally, at one point, in really I think like late September or early October of ’94, John Ludwig comes into my office and closes the door. He said, Ben, you know, I want you to take charge of this Web browser thing. I said, what about Phil? He said, well, don’t worry about Phil. Unfortunately, Phil kind of quit in a huff, and went to actually what’s now Real Networks.
That was kind of unfortunate, but he wasn’t making good progress, and we felt we needed to make more progress. I quickly got together a team of about seven people. Thomas Reardon was one of those folks. I picked up some people from the At Work team, which was sort of dissolving at that point. So I got Chris Franklin and Deepak Amin. There was a guy, Arthur Bloom, who was a new hire from college, program manager. There’s another guy, David Dickman, who’d been working on things like Internet shortcuts, for example. We had already started to do some of this integration with Windows. We spent October, November, December sketching out what the architecture and the plan would be for this Web browser, and what the features would be. Then Thomas spent some time with the MSN guys trying to license a code base, so we…because, if you can start with some existing thing that’s already a Web browser, then you can make it a lot better. You can understand the problems, especially we were able to license the Mosaic code, which was the browser at that point. All the compatibility issues are plain to see in the source code.
It actually turned out to be pretty nice source code. So we signed that deal in December 18th of ’94, and I gave the team a couple of weeks to obviously take off. Then started code base. Then we had a meeting on January 15th, where people were supposed to come with their development work items. Then we basically went, okay, who has the most important work item? We all agree, okay, that’s the most important one, and we wrote that person’s name down and wrote the task and the number of days. So we went around until we had basically eight weeks of work. So we gave ourselves eight weeks to write code, based on this starting code base. Then we drew the line--that’s all we’re going to do.
By the fall of ’94, actually I’m like…so the question is, okay, when’s Win 95 going to ship. That was always sort of our, you know, could we get done in time for Win 95. In October or November of ’94, we thought Win 95 shipped in February of ’95. So, okay, no, we can’t do that. But by January, it was like, oh, you know, Win 95’s not shipping for a little while still. It was more like June, July or August. Then I’m like, okay, we have a chance to make it. So we worked really hard, got all those things done, got a few more things done that we hadn’t planned to do, and got Explorer 1.0 done in time for Windows 95. We had already made this decision to put it in the Plus pack, and not in the Win 95 retail upgrade pack, for a variety of reasons. We didn’t know if it was going to get done in time, and packaging, art, and was the MSN service going to be able to support us with all those things. But we did get it in the Win 95 OEM release.
So every release of Win 95 that went to OEM customers, Compaq, IBM, etc., had IE as part of it. So the whole IE thing was fun. I basically worked on IE for two years, sort of August of ’94 to August of ’96. We shipped three releases. We shipped IE 2 in November of ’95 and IE 3 in August of ’96. I figure in those two years, about 17 months of those, I worked 80-hour weeks. Looking back on it, I go like, wow. It’s hard to imagine doing that again. But it was so fun. We were the underdog. It was fun technology. It’s new stuff, you know, it’s in the press. The Netscape guys are like we’re two years ahead of those Microsoft guys--they’ll never catch up. After a while, we’re a year ahead of those Microsoft guys--they’ll never catch up. Obviously, with IE 3, we caught up with them.
I remember there was an evening where myself and John Cordell, who I’d hired in November of ’94--he did a lot of the HTML and UI programming--and Chris Jones, who joined the IE team to do IE 3 with me. The three of us were…it’s like, I don’t know, midnight or something, we’re out at the basketball court shooting hoops. We’re like, what’s wrong with this picture? We’re all married and…what are we doing here? But it was a great time and there were a lot of challenges, a lot of new stuff to do. We got a lot of stuff done in two years.
LEE DIRKS: What did it feel like, putting that in, in Win 95, and seeing that go out? You shipped that product, and it was yours. What did that feel like, and then to watch it grow and develop through 2.0 and 3.0?
BEN SLIVKA: Yeah, getting IE 1 done was good. It didn’t have all the features that Netscape did. We were behind, and it was exciting and challenging, but it was also like, oh, my God, are we ever going to catch up? Navigator would like, they’d add tables, and they had Java script, and they had Java. It was like ooh, ooh, ooh. So you’re a little shell-shocked, but we had fought through…by May of ’95, we’d kind of figured out what was our IE 3 thing going to be. So I think…we at least had an idea of how we would catch up with them and get ahead of them. It really revolved around using COM--you know, the component object model, which we later…you know, with this whole Active X thing. But certainly, through IE 1 and IE 2, we were just totally in catch-up mode and behind and would we ever catch up. So, I think we were confident we could do better. But it wasn’t like we had won. There was no certainty there.
So for IE 3, we had a lot of these really aggressive things we wanted to do. We wanted to componentize the browser, so it was an HTML component that anyone could use. There was a component that any application, any Windows program could do HTP very easily. We had these scripting components and security and then the Java component. We did all this work furiously. Then there was this professional developers' conference that started March 12th. That was our rollout. We were trying to get a beta out, of IE 3. It turns out we decided to call it an alpha, because it wasn’t quite good enough. But that was when we turned the corner, and we got a lot of really positive feedback from all the attendees at the PC. Back on December 7th of ’95, there’d been this Internet strategy day, where we talked about what we were going to do, and Bill had said, made his little quote about Admiral Yamomoto and Pearl Harbor Day. But that was really where we kind of turned the corner, and could see that there was light at the end of the tunnel. We were actually going to be able to deliver better than Netscape. Then we worked hard for another five months, got a couple of betas out, and then shipped the final product in August of ’96. Yeah, it was a happenin’ time and a happenin’ place.
LEE DIRKS: I wanted to ask you…and I’m going to mention…during this period of time, there were a lot of memos that came out. Nathan’s Road Kill memo, Bill’s Sea Change, and then Internet Tidal Wave--how were those perceived? How were those taken? I mean, by the people within the company, whether in your group or outside of the group?
BEN SLIVKA: I think Bill’s memos were good. I don’t really remember the Nathan memo. But his didn’t, for some reason, get circulated as well. He tends to keep his not…his circulation list smaller. The Internet Tidal Wave was a good one. Paul Maritz did this Netscape as NetWare memo that was interesting. I think those were all great. It was like a call to action, a call to arms. Here, here’s this thing you should go focus on. People love that, whether at Microsoft or anyplace else--hey, here’s this thing to go do. You know, go put Internet into everything you do. So I think people liked that a lot.
I’m not a memo writer by nature, but I wrote this thing called The Web is the Next Platform. I worked on that for a couple of months and then released that at the end of May of ’95. A couple of days before, there was actually an Internet off-site that we did. There was an original Internet off-site in I guess April of ’94, which I wasn’t involved with. But this one in, it was like June 1st or something of ’95. So I wrote partly for that.
I really talked about how the Web is the next platform for applications, about how the productivity applications that we’re used to and focused on, like Word and Excel, that the new class of applications are a lot more content and service-based, and that the Web, as a platform, was a much better platform for delivering those solutions than WIN 32 APIs and GDI and WM Paint and all that stuff. I talked about--there was this thing called Blackbird. I said, hey, we should nuke that thing. I was just sort of like, hey, I’m just going to express opinions about lots of things. So I wrote all those things down. Included in that was kind of a one-paragraph summary, bullet points of what IE 3 would be. I said we’d deliver all of these in June of ’96. We were actually two months later than that. It was August of ’96. We did everything I had listed there and some stuff. We did Java, which I hadn’t listed, and we did these cascading style sheets and pics and a few other things. So we did a few more things.
LEE DIRKS: I remember the night of when IE 3 released. I mean, just thousands of downloads that particular night. That was when we really did turn the corner. Do you have any impressions about that, remembering seeing the reviews come in?
Internet Explorer 3.0 -- Success at Last
BEN SLIVKA: I don’t know, I was so tired at that point. You know fall of ’96 is sort of hazy for me. I think I slept a lot or something. But, yeah, we won almost all the reviews. If you looked at like the top 10 technical PC reviews, we won all of them--Mossburg and PC Computing, and InfoWorld and PC Week and all these guys. The only one I remember was negative was PC Magazine. Even they said…they said we had the best Win 95 browser--it only worked on Win 95. It didn’t even work on NT, because we’d done stuff where we relied on the Win 95 shell. But they gave the nod to Netscape, because Netscape had these cross-platform solutions. So we worked hard for two years and took on a leadership position. After IE 3’s release, then our market share started climbing. Before that, our market share in browsers was basically zero.
I guess another thing I should mention that’s important is that we got the AOL business. So, in…of course, all these things are subject to litigation right now, so I’m going to be a little careful of how I talk about these. But basically, AOL came to us in January of ’96 and said, we don’t want to be in the browser wars with you and Netscape. You guys are going to kill each other. We don’t want to be road kill there. We were in a great position to supply them technology, because we had this componentized browser, where they could host all our code and just have their own AOL frame over that. So AOL customers wouldn’t know the difference. In fact, there was this guy named Robert Sederman that writes this on-line newsletter, that was talking about it. He’d picked up the AOL 3 client and was using it, and he was like nothing’s changed. Then we went to a Web page with Java, and went, oh, well maybe there is some Microsoft technology there. But it was that seamless. So that was the case where, you know, Netscape was like, oh, yeah, AOL’s going to dive. It’s going to kill AOL. That’s what all the analysts are saying. We’re Netscape…we’re the browser leader. So they were sort of dismissive to AOL. Whereas, we were like, hey, we want to…you know, we have like zero market share--could you use our technology, please? So not only did we have a great attitude about helping them use Internet Explorer, but also because we were a componentized browser, we could do that, and Netscape was this big monolithic application. It used NFC. It was fat and slow. Anyway, so that was another I think important thing we did during IE.
LEE DIRKS: So you shipped IE 3, and then what were your next steps after that?
BEN SLIVKA: I think I rested up a little bit. Then the next thing was Brad Silverberg over in Java, I remember he asked me to take on the Java team. In retrospect, I don’t know if I would’ve done that, if I knew what was going to happen. But I was, what the heck, sure, I’ll go do that. After I’d been into that a couple of months, and looking at the contract and what was happening with the whole Java thing, I was like, you know, we’re going to get sued. I remember just thinking that. Oh, where’s the…there’s no way to avoid it, right? But, so basically we’d shipped a Java VM in IE 3, and that was people like Mike Tutongi(?) and Russ Arun and Victor Grabner and others who’d done an outstanding job…Cory Shrock(?) was the test manager. It won all the reviews. We had the best Java VM. It was the fastest, the most compatible. I built that team up, probably doubled the size of the team, and we did this great garbage collection stuff. We had this great just-in-time compiler that takes the Java byte codes and generates native code from them. We did a lot of work on our class libraries. So when we shipped the Java VM in IE 4, again, we won. PC Mag did a comparative review against Sun, Netscape and Symantec I think were the other three. We had the best Java VM. Again, it was the fastest by far, the most compatible--more compatible than the Sun VM, whatever that means--and the most reliable.
But there was a lot of controversy about…Sun was out talking about how Java was going to kill Windows. That was their game plan. They were very clear about that. There was confusion inside of Microsoft about what exactly we were supposed to do about that. Should we love it or hate it, and our perspective on the Java team was we loved the language, that Windows developers can choose Java if they want, to write Windows programs. So that’s the kind of thing we focused on. There was this sort of infatuation for a little bit with this thing we called AFC--application foundation classes--where we were trying to do sort of better cross-platform, write once, run anywhere, than Sun. We tried that for six months. It ended up being…we just couldn’t do it. It’s sort of funny. On the one hand, we’re out telling people, hey, write once, run anywhere, that doesn’t work. Then we were trying to do it ourselves. But it turned out we were right--we couldn’t do it. Anyway, so that was that little bit of stuff.
LEE DIRKS: And then you moved to the Windows user interface?
Neptune: Windows User Experience Prototype
BEN SLIVKA: Yeah, then I had my three-month sabbatical, or two-month sabbatical plus vacation. So I took three months off. I was building a new house. I got the house finished and took my family to France for a couple of weeks, and basically didn’t do any work for a while. Came back, and tried to…sort of sorted out what was going on.
Then we started up this Windows user experience prototype. The idea was to…if you think abut the Windows user interface you have today, you have the start menu and task bars and windows and close boxes and menus and dialogue, tab dialogue boxes, and tool bars and wizards and help and there’s a lot of UI stuff there.
So Steve Capps, who had come to Microsoft a couple of years before that, he was the developer on the Macintosh--the original Macintosh--for the finder for the first several releases. He’d gone on and spent seven and a half years working on Newton, which was Apple’s hand-held personal digital system. I worked on OS/2 for five and a half years--he worked on Newton for seven years. The idea was basically, okay, let’s throw all that UI stuff out and start from first principles, and just only take the barest elements we need. The central idea was let’s start over again with the user interface. Can we build up, given that processors are a lot faster, there’s a lot more memory, screens are a lot bigger and have all this rich color…you can do all these things like alpha blending and stuff like that…hard disks are a lot bigger. Can we build up a user experience that’s both simpler for novices or people that’d never used computers, intermediate people, but also scales to be more powerful for power users? So that was kind of a neat thing. So we’ve been spending the last year working on that. You can see the Web site, if you want. It’s on the internal Web, Neptune, to get more details about that.
LEE DIRKS: Can you talk about what the organization looked like in 1985, when you came, and what it looks like now, and how it’s changed over the time that you’ve been here, what you’ve seen and what you’ve witnessed?
BEN SLIVKA: Sure. I think, when I started, there were maybe about 800 people at Microsoft. We were in the Northup Building down near 520, next to the Burgermaster. Then, across 520, there were some other buildings; the east campus I think we called it. I can’t remember what the name was. So I was on the second floor of the Northup Building at sort of the southeast wing I guess or something. I remember just around the way were the executive offices, so Ballmer and Bill Gates and all those guys were down there.
I was hired into the DOS team, and it was really like the systems group. I guess there might have been, I don’t know, 20 people or something like that. There were three people working on MS Net and these people working on UNIX clone called XENIX, and then this OS team. Jon Shirley was president. We had one recruiter, I think, maybe two. Quite a bit smaller than we are today. We didn’t have a LAN. The software development was done on a…we actually had…people’s desktop machines were either PC XTs, which was an 8086, no, sorry, 8088, running 4.77 megahertz, and a 10 megabyte hard drive. Or some people were just getting ATs--IBM ATs--which, I got one of those, and my next-door neighbor was pissed at me because I had one of the new machines. That was a six megahertz 286, again with about 640 K of RAM, and a whopping 20-megabyte hard disk, I think.
The way we did source control was there was this minicomputer thing, and you had a 19.2 serial line, and you did uploads and downloads with FTP or something and ran SCSS on the server, to do source control. So I thought that was a little silly. So I think I strung maybe even the first LAN at Microsoft. So I was like, well, there’s this LAN stuff, can we put a network together? So I was working for Mark Zubokowski--he was my first manager. He was one of the two or three guys that worked on MS DOS. I like figured out what cables I wanted and ordered net cards and got some contractors, or whatever, to lay the cables. I put the first two servers up, which one was Hagar and the other one was Opus for the cartoon characters. These were 286 machines with 20-mg hard disks. Those were our file servers. Then we had a way to synchronize that source code, so people could get the source code off of those. It was a little bare-bones kind of, at that time.
What else? I remember the first day I was at Microsoft. Actually, I went over to Burgermaster, got a burger and sat down outside of the…just on the grass at the Northup Building. I was just eating my lunch and going, hmm, I wonder what the future’s going to be like. Not really going like was this the right choice, but just, you know, you have sort of these little butterflies or willies, that when you start something new, how’s it going to turn out? I remember that first day.
LEE DIRKS: So you had no idea.
BEN SLIVKA: No, idea, right, no idea I’d be here today talking to you.
LEE DIRKS: Can you define Microsoft culture? I mean can you give me some ideas or some impressions of what you think Microsoft culture is, and how it might have changed over time?
BEN SLIVKA: I think smart, focused on software. When we think about recruiting, and I spend a fair bit of time on that, we’re trying to find people who are very smart and very passionate about software. That’s a continuous thread that runs from, when I started at Microsoft, it was here before I came and continues to this day. So I contrasted that to when I was at IBM for a year, ’82 to ’83, I remember the first week I was there, I sat down with my manager, this guy, Alan Atherton I think his name was. He’d been like a fifth-level manager someplace at IBM and then screwed something up. He was kicked down. He was only a second-level manager. He sat me down in his office and showed me these hand-written notes from when I’d interviewed there, several months before. Smart and he knows it, aggressive, you know all these little adjectives and comments about me, and he said, you know, Ben, you should tone it down a little bit. Right there, I went, oh, okay.
The perception I had of people at IBM is that they, the managers, they were just trying to cover their asses. They didn’t want to take a risk and build some really great software. They just wanted to make sure they didn’t screw up. I think Microsoft’s all about taking risks, making big bets, and trying to find what are those big growth markets and then going after them really aggressively.
We’re larger now, and so I think there are probably pockets…I’m sure there are probably places inside Microsoft and go, ooh, that feels like IBM to me a little bit. I think that’s a problem. When I see places like that, I try to do something about that. It’s not always easy, but I think that focus on really understanding what the problems are. The focusing is really important. There’s a vast space of things we might attack, of problems we might attack, of customers we might go after. It’s important to focus on the things that you think will be the highest return. So I think we’ve been pretty successful at doing that so far.
LEE DIRKS: What has been your most rewarding, or most challenging position that you’ve held at Microsoft?
BEN SLIVKA: Rewarding, challenging…I wonder if those would be the same, or not. I think doing the IE project for two years was the most rewarding. It was great fun. I got to build up a team myself. I went from, sort of grabbed these six or seven people quickly, and then, when I did IE 3, I really staffed the team up. I hired like 35 people in the space of about four months. By the time we shipped IE, I had like 60-some people working for me. So at that point I had the largest group I'd had ever. So just the excitement of the technology, of the competition, of hiring all those great people and working with all those great people, and turning them loose to be successful and the learning opportunities that they had. I certainly had some learning opportunities, too. So I think that would be certainly the most rewarding time. Although, you know, MS DOS 6 was a lot of fun, too, but a smaller project.
The challenging time, hmm, well, there’s probably lots of those. OS/2 was sort of a very frustrating and challenging time. It was sort of hard to overcome a lot of things. The WIN 32 for DOS thing -- I didn’t make sure it was clearly enough articulated what it was we were trying to do. That was pretty challenging. Things where I was actually successful, those weren’t challenging, because I was successful, so I figured it out.
I think the phase we’re in right now, where we have these really successful products in Windows, in Office. We should talk about this middle-management retreat I put together. I think that would be a good thing to talk about now.
I’m on sabbatical and it’s July or August. I’m coming back in September of ’97. Natalie Yount, when I first met her, was the librarian at Microsoft. My mom was actually a librarian here in Seattle, and she knew Natalie. So, actually, when I was here to interview that February day in ’85, I said hi to Natalie. She calls me up and now she’s in the Human Resources group and sort of responsible for employee retention or something. I’m never really sure exactly what her job is. She says there’s this power lab thing in Massachusetts I want you to go to for a week. I’m like, really? Do you really want me to? Oh, okay. I had no excuse. There was no work I had to do. So, I’ll go.
She gets 14 of us signed up and we show up at this place. There are…it’s basically what they call immersive social experience. I’m not supposed to give you many details, because it sort of spoils it if you want to go do this thing. But it’s not like a cult thing or anything like that. It’s this immersive social experience, where there basically are these tops and middles and bottoms, and you get kind of thrown into it, and you see what happens. It’s basically sort of a little mini-society, and you see how that evolves over a period of days. So that was…as we’re doing that, we’d take these timeouts, where we’d get a little sort of theory about what’s going on and what happens. Eventually, it ends, and then there’s another day or so of some other exercises to learn more about that.
It’s put on by this group called Power in Systems, and Barry Oshrey(?)’s the key guy there. They’ve been doing this thing for 25 years. They do it once or twice a year. He’s basically got a model of how power in systems of people, how it manifests itself, and what the various conditions were. So it was interesting to come back to Microsoft after having done that and see that a lot of sort of classical problems that this guy identifies and situations and all that, that those were happening at Microsoft.
Sometimes I think at Microsoft we have a little hubris about, okay, well, we’re the largest market-cap company in the world now, trading off and on with GE. We’ve been super successful. Bill’s the richest guy in the world. People go, wow, we’re pretty "studly" at Microsoft. We know everything. We never make mistakes. And we’re different. Rules or behaviors or guidance, or the way things work at other places--that doesn’t really apply to us. I think we have some hubris about that. So I was interested to come back from this power lab and go, hey, whoa, that’s all happening here.
I’ll kind of give you the brief sketch of his stuff. Basically, there are tops and middles and bottoms. The bottoms like have no idea about what’s going on above them. They’re just like, random stuff comes down and they’re like, whoa, we have no idea what’s going on. The tops are like, they’re living in all this complexity, trying to make all these decisions, and they’re out of touch with kind of what’s going on below. Then you’ve got the middles, who are trying to keep the tops happy and keep the bottoms happy. So they’re in the middle. Then they tend to like struggle with one another. So a lot of that stuff was happening. So there are sort of prescriptions or recipes about what do you do about that. The tops have to communicate regularly and clearly about what’s going on, and the bottoms, somebody’s going to have to give them feedback. The middles have to, instead of fighting and warring with each other, they have to kind of integrate, how can we help each other.
So I got back from this thing in September and said, you know, we have a problem here we should do something about. So the same day I got back from a plane flight, I sent a mail to Paul on it. We have all these problems. I think we should have a middle management retreat, you know, get some middles together and talk about it and learn about this, and see what we can do.
So then I eventually got hooked up with Natalie Yount, and then this guy, Kevin Purcell in HR. So I put this thing together and we had it in November, a couple of months later. There was actually a lot of resistance. Natalie was like, don’t go out on a limb, Ben. Don’t just go do this by yourself. Get other people. Other people are like, slow down. But I was convinced that we should just do this thing, because something good would happen. I didn’t know what, and people are like, well, Ben, you have to have a reason you’re doing it, or like what’s the outcome. And I don’t know, but let’s just do the thing. Let’s get the date. So we picked the day, we got people invited, and then I actually got several people to sort of co-sponsor it with me. We talked about how we would organize the day. It was actually pretty good. So we got together and talked about this stuff.
There was…you know, one of the things you’re not supposed to do is whine. Don’t just complain about stuff--offer up a solution, right? So one guy actually put together like no-whining shirts. No whining with the little red slash through it--Anthony Bay did that. So that was a pretty good thing.
The biggest thing, I think, a thing I’ve seen is that we’ve grown very rapidly and there aren’t…we don’t have a lot of role models internally about how do you manage a large organization. People like Brad Silverberg and Jim Allchin and Paul Maritz and Steve Ballmer and Bill Gates, I mean they’re learning on the fly. How do you manage an organization this large? So one thing I think is that people like Bill and Steve and Paul and others--they have to…their job changes. They don’t get to do the things they’ve done in the past about doing detailed product reviews and technical evaluations and smoothing all that.
They have to take on much more of a coaching role, and think more broadly about managing the organization that way. So, actually coming out of that MMR thing, we had a follow-up thing that Paul Gross organized in March of this year. Then Paul Maritz actually sort of took this over and renamed it the Senior Leaders Forum. He thought the Middle Management Retreat was too, I don’t know, something. He actually got a very similar group of people together recently and we had a day and a half off site to talk about stuff. So that was a good kind of, having done that MMR thing, it seems like we’re going to be doing that kind of stuff.
But it’s a real challenge. When I look at the…when I think about the future, nobody knows what the future is. People have all sorts of ideas about that, but you can look at our strategic bets we’ve made in the past. The GUI thing, that was a good bet, you know. Now, to be fair, we spent all that time on OS/2 and Windows was just, oh, this is just little side thing, right? The ratio of investment of people on OS/2 versus Windows was five to one or 10 to one, I mean we had way more people working on OS/2. But Windows obviously was the thing that was successful.
Similarly, you can look at like interactive TV and the Internet. You know, we have all these people, 200-some people working on interactive TV, and a few people working in the Internet, doing the plumbing stuff. Eventually, that got organized the right way. So the thing that’s important for us to understand is that the future’s a complicated place. I sort of devolved the mathematics a little bit, that, if you think about the future as this sort of multi-dimensional space, and you want to identify the basis vectors…you know, what are all the things like, for example, high bandwidth connectivity in the home. Does that happen soon, or not so soon, or like a long time away? The people in homes, do they want to have a lot of devices, or do they just want to have one? Are they connected in a wireless fashion, or in a wired fashion? How quickly will people do like home banking and stuff like that? How quickly will not only consumers adopt that but will people who bill you be able to do that kind of stuff? What about the TV? Is a computing device with the TV, is that a good thing, or not a good thing? That’s just the consumer space. You can think about the enterprise space, as well.
So I think we really need to think about what are those basis vectors--all those sort of variables that you have to consider. Then look at the multi-dimensional space and pick some areas there that we should really be going and attacking. We’re…I think certainly, I think in the Windows group we’re being a little to focused on this whole NT thing. I think we need to widen our thinking a little bit there. I also think about what’s our new OS effort? As long as I’ve been here, that’s been 13½ years almost, there’s always some next new operating system we’re working on. At this point, there isn’t one. The research guys aren’t really doing one. We’ve got this NT thing. I’ve just going like, is NT going to last forever--I somehow don’t think so.
A particular thing I think is…there’s this fundamental shift that’s happened now. Certainly when I started in ’85, we were focused really on these productivity applications. It was about building a platform for word processors and spreadsheets and drafting programs. It was features, features, features. Add more features to those things. Well, you know, we have enough features now. You could always add more, but I think that…when the Office guys say that the bulk of the feature requests they get are for features that are already there, that people can’t discover.
Microsoft as a Service Business
So what I think about is GUI was the watchword for what we did these last 10 or so years. I think, going forward, another watchword we could have is service. Let me describe a little about what that might be. Today, when we write software, we think about features first, and then maybe we think about performance, then reliability and all that other stuff is kind of do as much as you can. Of course, people are very frustrated with our software now. You know, it’s not reliable. It’s very fragile. It’s really complicated. So, what if we said now our number one goal is to provide great service, and we’re going to sell you Windows and that’s the Windows service, and we keep it up to date and we make sure it works for you?
So the interesting thing to say is, let’s start a new operating systems group and, from the ground up, write everything around a code assuming that it never gets to reboot. It never gets to crash. It never gets to give the user an error dialogue that the user doesn’t know what to do about. It constantly keeps itself in tune. So think about an operating system where it assumes it’s connected, and so when it needs a new device driver, you can go fetch one off the…even if it has to dial the phone line, but at least that’s the way it gets it. Think about every time it has an error or something, it reports back to somebody, so that stuff’s centrally collected and you can go, what are people having problems with?
A company like AOL does that today. They have their basic software, their AOL client. They upgrade it on a regular basis. It reports back problems that their users encounter. There’s all sorts of usage statistics about what people really do. We’re like, we ship the bits out, and then we have no idea what happens until Walt Mossburg writes a column, or Michael Miller in PC Magazine writes a column about something. That’s a very random, haphazard way to get feedback. So we don’t have any regular feedback mechanism. You can imagine doing all these things on top of Windows. Like take our existing…you know, NT 5 is what, 40 million lines of code...and somehow wrap things around it to kind of understand these things. I think we should do those things, too, but it is interesting to why shouldn’t we have a new operating system effort.
LEE DIRKS:What elements do you feel have contributed to Microsoft’s success?
Elements Contributing to Microsoft's Success
BEN SLIVKA: Okay. Well, it’s interesting to think about companies in general. Why do companies fail and why do companies succeed? Companies are just people, right? So I think, you know, Bill has to be a very important factor in the success of Microsoft. He and Paul Allen had this vision of a computer on every desk, in every home, and they pursued that tenaciously. They said they were going to be a software company at a time when most people are deciding to be hardware companies.
Bill’s real focus on hiring really smart people who are really passionate about software and getting people like Steve Ballmer and Brad Silverberg and Paul Maritz, and I mean all the great people we’ve hired. That focus on excellence of the people has been very important. Then, our focus on software and thinking about how can we create a lot of value in high-margin ways…if you look at…you know, what’s the rap on Microsoft? We’re not inventive, innovative. We steal ideas from people. But the reality is, we think about, okay, what are the most pressing needs that people have that computers, as they exist today and as they might exist in the future, can solve in a pragmatic way?
Now, we’ve had things like Microsoft At Work. Let’s put Windows 3.1 into a phone--not very pragmatic. That didn’t succeed. We tried this Bob thing, which had interesting ideas about characters, but it was too slow and it didn’t scale up. You could do simple things with it, but then you ran into this brick wall. You couldn’t do anything past the simple stuff.
That kind of focus on what are the really high return things that we can go do, where PCs can make a difference. So some people talk about, oh, yeah, people use PCs to manage their recipes at home, or home control. There’s lots of things you can possibly imagine doing with PCs, or with computing in general. We’ve been, depending on how you want to think about it, lucky or smart or whatever, and we’ve picked things like Windows, like Office that were sweet spots, that really solved some big problems for people in compelling ways. Then we just executed that better than other companies. So, if you look at…you know, we started out as a little small company and there was big IBM and big DEC and big Wang, and there were a lot of other companies. The size doesn’t matter. The past successes don’t matter. How much money you have in the bank doesn’t matter. This whole DOJ thing is sort of ludicrous about all of those…you know, oh, my God, Microsoft has all these things, a monopoly position in the operating systems.
That thing is so tenuous that…you know, our software is so fragile today and so complicated today. Someone else could come out with something simpler, more reliable, and I think customers would love to have that instead of Windows. So the question was about sort of why we’ve been successful. So it’s really about…in the end, it’s just about people. It’s about having the right leadership and then hiring great people everywhere. If you don’t hire great people, you’re on a path for destruction.
LEE DIRKS: You mentioned a second ago that Microsoft gets dinged for not being innovative. Can you talk about, or come up with some examples, in your opinion, where Microsoft has been innovative?
BEN SLIVKA: Sure. Well, there’s lots of innovations we’ve made. You could argue that our business model itself has been an innovation, in that we said we’re going to be a software company. We’re going to provide operating systems. We’re not going to be a hardware company. We’re going to help grow the PC industry by helping provide an environment where the PC manufacturers can go knock their brains out competing with each other on PCs. Hardware manufacturers, whether it’s video card manufacturers, or speaker manufacturers, or any of those things, they can go compete against each other head-on…the printer manufacturers and software vendors. I think the term is horizontal integration, or something, instead of vertical integration. So that business model thing--that was a very important thing we did. Macintosh didn’t do that. You know, nobody else has done that. That’s been why a company like Dell, for example…Dell’s stock over the last 10 years has outperformed Microsoft’s stock, which has done pretty darn well.
If you look on the technology side…you know, we have this rap about, oh, not the innovative technology. But you have to look at the whole picture. In the end, it doesn’t matter how "innovative" a piece of technology is, if people can’t use it, if it can’t be manufactured in volume. The fact that something is successful is because it’s innovative and it’s made the right set of innovations. I mean, if you look at the browser stuff as an example, we used our component object model, which was work that was started in the, I think the early ‘90s, maybe even the late ‘80s, to build component software in a way that I don’t think anyone had done before. We actually had eight or 10 meaty, heavyweight components, you know, a scripting engine, an HTML rendering engine, a protocol engine for HTP, FTP and Gopher, a security component, a Java component, a document hosting component. That let us get our work done a lot more quickly, I think. If I really think back on like why did we beat Netscape, one of the reasons is we’d chosen to componentize. It let us have these teams build these independent components and then glue them together in a reliable way. So, doing Active X hosting, doing the security stuff we did there…we’ve done little trivial things like toolbars, or whatever. So I think that’s a good example.
The work we did on Windows NT, if you wanted to be mean about it, I guess you could say Windows NT was just sort of a VMS clone or a minicomputer clone. But our key thing we did there was we built an operating system that, while today not quite reliable enough for some uses, is pretty darn reliable. It’s a much easier operating system to administer than UNIX is, for example. You know, the UNIX system admins are like have to know all these esoteric command line strings and Windows ’95, you can manage it and just do these little GUI things. So that lowers the cost of ownership for certainly enterprise customers and other customers.
Some of these things are sort of mundane, right? But the fact is they actually deliver real value to end users. So the popularity of these things, if you look at Office, Excel, first on the Macintosh…some people…there are all these myths about Microsoft. Oh, Microsoft applications are successful only because Microsoft owns Windows. Well, if you look at the Macintosh, where we were first successful with our applications with Excel, with Word, we didn’t have access to the operating system, I think. It’s pretty clear that Apple was slightly hostile toward us about that. Yet, we delivered GUI applications that were better than anybody else did on the Macintosh.
We took that expertise about how to build GUI applications and did that same work on Windows and did it better. So Excel was a better product than 1-2-3. It simply was easier to use, was more powerful, had more features. So we won that market fair and square. 1-2-3 was the market leader in the character mode, spreadsheet market. WordPerfect was in word processing. We actually looked at Office and we said, let’s take these applications that work sort of differently and try to make them more uniform and consistent, so that people who do…and a lot of people do spreadsheets and word processors…and presentations, and try to make those things more similar. So I think that was another innovation. So those are some examples.
LEE DIRKS: Very good. Can you talk about the milestones or key events that you see in your opinion that have happened in the time that you’ve been here?
BEN SLIVKA: Sure. Summer of ’85, I think shipping Windows 1.0 in November of ’85 was a milestone. I think, first the deal with IBM on OS/2, which happened that Fall, I think, of ’85, and then the eventual divorce from IBM were certainly milestones.
Really, the divorce from IBM was when we said, you know, we can stand on our own. Before that, there was this idea that somehow we needed to have a single key OEM partner. At some point, that may be…that was probably true at one point. At some point, that became untrue. I don’t know when, but certainly we divorced IBM, I think that was true. So I think that was a real statement, we’re our own company and we’re going to control our own destiny. I think that was a very important milestone.
I think the release of Windows Excel in I think ’87 was important, as the first real Windows application for Microsoft, and it was a very impressive application at that time. Windows 3.0 in April or May of 1990, where Windows really was now a platform sufficiently rich enough and capable enough that ISVs could really target that.
That was a very important step forward, because there was no certainty that Windows would be in the market prominence that it has today, back then. I mean it was DeskView, it was IBM, it was GEM from DR DOS. There were a lot of people trying to establish the GUI platform for PCs. Macintosh was still absolutely a threat. You know, Macintosh never got…they never figured out that if they licensed their software and let the free market kind of turn the hardware crank, that that would be a good thing for them. They just never figured that out, even though we told them, and lots of people told them.
I think Windows 3.1 in ’92, May of ‘92--that was the release of Windows where really corporations and people could really start deploying that in volume. There were enough applications. That was a pretty critical point in time. I think certainly Windows ’95, in August of ’95, where we got people over to 32 bits and got a nicer user interface, was very important. Then, somewhere there, we shipped the first release of NT. We kind of got that stake in the ground--hey, we’re going to go get serious about the enterprise. Those were important things to do. The, I think some of the IE stuff, IE 3 maybe--I don’t want to over-promote anything I worked on, but I think IE 3 probably was a good point. Looking forward, I think…and there’s some milestones you could pick with NT 4, I suppose. But, going forward, I think NT 5 will be a big milestone. I think SQL 7 will be a big milestone against companies like Oracle and Sybase and Informix.
At some point, we’ll get our on-line service act together. You know, we’ve got things like CarPoint and Expedia, Home Advisor…that are nice. But MSN really hasn’t captured the imagination of consumers at home or other folks. It’s definitely second-fiddle to AOL. I think we have to do something there.
There’s certainly a possible future where you don’t get to charge for software bits anymore--it’s all the service that people are paying for. Just like we have TV cameras and monitors around us now. Those aren’t the people that make the big bucks, right? It’s the TV networks. I don’t think that, in the future, if you go out 10 or 20 years, it’s not entertainment programming, which is where all the value is. But it’s things like home banking, and managing your mail, and scheduling things, and just all the ways that computing will help people use their time more efficiently and enjoy their life more. I think there are a lot of services aspects of that. There’s someplace where Bill said this at some point, that in ’75 he and Paul Allen thought that software would be more valuable than hardware--that that would be the biggest expense of a computing device. I go one step further and I think that at some point the service will be the thing that people pay for and the hardware and software costs will eventually be zero. You know, if you think about cell phones as an example, that you can get a cell phone for zero. There’s software in there, right? And there’s a piece of hardware there. But it’s the service that’s…because that’s where there’s this ongoing revenue stream.
LEE DIRKS: You talked a little bit earlier about the importance at Microsoft of making the right hire, doing good recruiting. Can you talk a little bit about development, management, leadership and the role that those actions play at Microsoft?
Development, Management and Leadership at Microsoft
BEN SLIVKA: Sure. The thing that I’ve come to realize over the years is that it’s very important for me when I’m building a team, to make sure I’ve got the right person in the right job. Overall, for Microsoft, we’re more successful when we have the right people in the right jobs, because they’re more motivated. They’re better skilled for those things. Now, the next step then is, okay, we hire a lot of people off college campuses. How do we sort of promulgate our culture, and how do we make sure that people are growing and learning and becoming more and more capable and more valuable to Microsoft, and to themselves? That’s a thing, which I don’t think we’ve been very good at.
We’ve grown so fast, and we do have a lot of people at Microsoft who’ve never worked anywhere else, or they’ve worked someplace else for a little bit of time. So by contrast, I was asked to join the Board of Trustees at Northwestern recently, which is very educational, by the way. But the President who’s at Northwestern now, Henry Bienen, had been at Princeton for many years. The President of Princeton had grown all these people. There were 10 or 12 university presidents around the country that worked for this guy, or underneath him somehow. So there was a whole culture and system and way for people to be grown into leadership positions. We are so busy with building products and the strategy of the day and the competitive threat of the day, that I don’t think we spend enough time thinking about how we grow our people. It’s not that we don’t want to, but it always seems to take sort of second priority. So I think we’ve been a lot more haphazard about that than we could…now, to be fair, there are examples of companies that are very methodical about that. I read about Jack Welch, who is the CEO at GE, and they have dossiers on the top 3,000 people at GE. He apparently reviews them annually, all 3,000, and they talk about what, where is that person going next, and really thinks through all those things.
I don’t know, maybe that’s too mechanical, and maybe…the thing that’s great about Microsoft is we hire people who are passionate about software and are smart. We’re not trying to hire people who are just sort of political, get-ahead, climb the corporate ladder thing. Although, we’re large enough, we certainly have some people like that. But I think there’s some balance in there, where we should think about who are the top people. Let’s make sure they’re like moving around, like maybe people from Office should go work on Windows, and people from Windows should go work on Office, and we should shuffle people around I think maybe a little bit more than we do. I think it would be good for those individuals certainly, and also good for the groups that they join.
Also, you know, managing people is something I’ve haven't always paid attention to. I probably first had people working for me after I’d been here a couple of years. After a while I got to realize it’s my job to make sure that they’re growing and having learning opportunities. So that’s important. One of the challenges we have as a company is that many of the people who have been here several years, you know, that stock price certainly went up. So a lot of people don’t have to be working at Microsoft, don’t have to be working period. I think I probably fall into that category.
So we have to make sure we’re continuing to provide challenges and excitement and intellectual and emotional rewards to those people, because the monetary stuff isn’t going to cut it. So in some ways I think that’s our biggest challenge. I know I’ve heard Bill say several times that our biggest challenge is how do we keep the great people that we have, and keep them motivated, excited. It’s harder for us, because of that tremendous stock appreciation. You’ve got a company like GE or other places where you have people that can sort of focus on what’s my corporate ladder…kind of thing. But I think I’d rather be here than at GE. It’s not the money, it’s because people aren’t so concerned with the money. They’re more concerned with are we building the right products, are we satisfying customers, are we doing exciting work. But I think that is a big challenge, to try to focus on growing people a little bit more.
LEE DIRKS: How do you feel, looking back, on 13-plus years at Microsoft? I mean, did you ever think you’d be here 13 years?
On Working at Microsoft for 13 Years
BEN SLIVKA: Let’s see, well, some days I’m tired. But I’m not doing 80-hour weeks right now. But I might start again pretty soon. Some people really plan ahead, and they have this list of a hundred things they want to do before they die. I’m not that way. You know, some people apply to nine universities and I applied to Northwestern because that seemed like kind of a fun school. Then MIT and Cal Tech or something were the other two schools. I decided I wanted to go to Northwestern. So, luckily, they admitted me. I’m not much of a planning-ahead kind of guy. I think, oh, what seems to be a good thing, and then I just go do that. So I don’t even think the thought crossed my mind, how long would I be at Microsoft. I was just like, oh, this seems like a good thing to do now. I look back on it and go, you know 13 years is a long time. I look back…I wrote this CV for Northwestern--they really wanted me to come on their Board. Oh, I can do a lot of little projects here. It’s sort of funny.
If you’d asked me…five or 10 years ago, I was kind of like I don’t really know that much, but, after a year or two, all of a sudden, I turned around. I go, whoa…I’ll be in meetings or something…I just kind of get this sort of feeling, oh, I’ve got a little bit of experience here now. It’s sort of funny, because, when I was at IBM, the people who had "experience", they were like, they didn’t seem very clever to me. So I always thought that experience, that doesn’t really count for very much. But now I kind of go, oh, so the experience is like all those mistakes he made, right? So, yes, some days I feel pretty old, I guess, and other days, I go, wow, there’s so much more to do, that it’s sort of like my first day on the job I guess in some ways.
LEE DIRKS: Were there any interesting stories, funny incidents, embarrassing moments, for yourself or others?
BEN SLIVKA: I should have thought about that beforehand. There’s those classic ones, where there’s people jumping in lakes and all that. But that’s kind of, those are kind of silly. I remember when we shipped Windows 95, maybe I could tell this story. So we shipped Win 95 on July 14th, so there was a little party. You know, there was like this signoff ceremony and then a party in the afternoon. There were tents outside of Building 5. I think I…I started this tradition with MS DOS 6 release. My wife made some really incredibly, thick, rich, brownie, chocolate things, and then I’d bring some nice bottles of champagne, because we like wine. We’d have little sort of mini-parties in my room, and people would come and have champagne and eat these really incredibly-thick brownie things. So we had a little champagne that way.
Then we went out and Silverberg or Cole had bought a couple of cases of Dom Perignon Champagne, which is not my favorite champagne, although it is pricey and James Bond drinks it. So people were drinking and spraying things on each other. So then we went over to the fountain there at Building 8, to kind of clean off. We got into the fountain, and some people had had more to drink than other people. In the fountain, you can’t really see below the surface, but there’s all this maze of these pipes. So some people got all bloody and wet. John Ludwig, who I worked with at the time, got all bloody, on his knees. He came in the next day and he was like very gruesome looking. I got all wet, and I remember…oh, yes, so I was taking off some of my clothes there. So I called my wife to pick me up, because I figured I shouldn’t drive home. Although I hadn’t had that much to drink, but I still wanted her to drive. I remember plodding out through Building whatever with just my underwear on.
What other things did we do personally, embarrassing things? You know, usually these things happen around ship party thing. I remember when I…the Stac trial was down in L.A., downtown L.A. and next to I think they call it Japantown, or something like that. There are a lot of Japanese restaurants around there. So I’m there for two weeks, and I’ve got the lawyers there. I like to eat Japanese food, so we’re going to go try these different Japanese restaurants every night. So I was…my goal was to sort of gross out these attorneys basically, so I ordered really obscure things. What’s the…there’s that particular…oh, sea urchin--I’m forgetting what the Japanese name is, but…so I had that, and I kind of grossed them out with that and some other dish. I don’t know, we’ll keep going. Maybe I’ll, something will occur to me.
LEE DIRKS: That brings me to the end of the questions that I had for you. I wanted to give you a chance, if there were any final thoughts or bits of wisdom.
Final Thoughts -- Microsoft's Future
BEN SLIVKA: Okay…wisdom--well, I don’t know about wisdom, but…well, I kind of got a few things in. Microsoft’s been very successful. It’s been because we’ve focused on customers and doing things that we can sell in high volume. Now, I think we’re at a crossroads right now where our customers are changing a little bit. It’s no longer spreadsheet logs(?) and word processing geeks--those aren’t the people that, going forward, is the big growth market. Those people are always going to be there. A lot of us are those people--we use PCs as an everyday part of our lives. Without e-mail we’d mostly be dead in the water I think, here at Microsoft.
But going forward, we have these enterprise customers, where there’s all this business to go get there that IBM and Oracle and others have today. Their demands are much more around things like reliability, availability--like the system’s always up, serviceability, if something goes wrong, you can fix it. IBM used to call it…when I used to work at IBM in HR, I remember RAS(?), or maybe that’s when I worked on OS/2 with IBM. It’s lost to antiquity now. So that’s the thing that the enterprise customers care a lot about.
When you think about people at home using computing, that same service thing applies. But it’s just got to be a very reliable thing. So I think there’s just a big shift we have to make as a company from features, you know, more and more and more features, to providing great service. I think there’s…as I said, software’s a service. We move from selling bits that people install on machines to providing an ongoing service. I think that’s the thing that will happen. So there’s a lot of cultural implications to that. How do we organize the product teams around providing great service? Right now, we’ve got the product teams in all their little silos, right? The Windows team and the Office team, then we’ve got this IMG team, MSN over here.
Separately, there’s this service group, PSS over here, like Microsoft technical support. We have all our little silos. If, by contrast, you looked at a company like AOL who’s delivering a service today, they have a much more integrated view of that, and that whole customer experience, and how can we deliver a great end user experience. So I think that’s the change we have to make, where it’s not how many features you add, but it’s how many error messages did you take out of the product. How few lines of code do you have? How many times you don’t change the code, so that things can stay stable and reliable? Try to be clever about that.
I think the second thing is how we manage the company. I described the future, and tried to factor in this multi-dimensional space, and then pick these bets. I think there are…as I say, that’s today…we’re not making like where’s our new OS bet. Where are the things, where we can get just really blindsided? I think there’s some things we should go after there.
Really just thinking about, in general it’s computing. How do we make computing a part of people’s everyday lives? Obviously, for a lot of us today, without a computer, that would be pretty hard. But computing, because I’m not trying to focus on PCs or software in general, but the idea of applying computational power to information, and how people communicate and shop and interact with one another. Really, to think that through in the home and in businesses…and I think the Web lifestyle vision that Bill has, the digital nervous system, which is this cold, chilling thing--those are a start at those kinds of things. I don’t know if those are the best ways to express those or those are the most compelling things. Maybe it’s just to flesh out underneath what the implications of those things are. But, that idea of bringing computing to people’s lives to help them use their time more efficiently. There was a Wired Magazine article a year or two ago. It talked about the scarce resource is not a physical one. It’s not grains of sand or gasoline or food or anything else. The scarce resource is people’s attention, so figuring out how to optimize for that. Those are the things I would think about.
LEE DIRKS: Fantastic. Well, thank you very much for coming out today. Maybe we’ll touch base again in a little…couple of years, and see what else has developed.
BEN SLIVKA: Come back in five years and see if I was right about service.
LEE DIRKS: Exactly. Right, that’s very much.
BEN SLIVKA: Thank you.