A Window to Gates

A Window to Gates

Microsoft: Windows apps a safe choice

(Microsoft Chmn Bill Gates)(interview) (Cover Story)

John Desmond
Software Magazine

September 1, 1991

Q. What do you see as the role of the mainframe--dinosaur or data server?

A. A data server. You're going to continue to have a hierarchy of computers: enterprise-level computer, a department-level server and the desktop machine. In the information age where so much more information is going to be in electronic form, the demands on the enterprise machine and the department machine will just go up and up. That's why we're making our operating systems run on so-called multiprocessing servers.

The mainframe will have to compete with other devices for that role as the enterprise machine. It has a huge advantage because it's already there and it runs a lot of software. But as you add new databases and advanced image applications, or taking paper information and putting it online, you'll look not only at mainframes, but also other architectures to provide that new capacity. This NCR 3000 series machine is an example. It's kind of like a mainframe, but it uses lots of microprocessors. So do you call that a mainframe or not? It runs Microsoft-type operating systems as well as Unix. So it's within our domain. It's something that we'll write software for.

So the role of the central server, if you think of it in that sense, the demand for capacity and cycles is going up. So people who understand about mainframe issues of security and database access are in a strong position. The people who make the hardware are going to find it immensely competitive.

Q. As an operating system development company, how do you view IBM's MVS, which was described to me by one of its developers as one of the "crown jewels of mankind"?

A. It's an inspiration to people who write software to have a system they can keep raising the price on year after year, and customers are very happy about that. So the fact that you can sell software for that kind of money is impressive. People must value MVS very, very highly.

I think in terms of the source code; like all systems that have been around a long time, it gets to be fairly large and complex. When we have systems like that every four years or so we start a from-scratch project to do a refresh, and have compatibility. I'm not at all familiar with how much they've done that with MVS. They may have done it a great deal. I've used MVS, but I don't know its internals at all. I know more about its pricing than I do about its internal design.

But crown jewels of mankind? I don't know what that's referring to.

Q. How about DB2? Is that something you'd like to emulate someday on a desktop level?

A. IBM on their mainframes has proven that it's very hard for third parties to provide a database. DB2 has been very successful. It's a good product. A big issue for a lot of the customers that we're working with is moving the data into relational form and moving the front end into graphical form.

When you move the data into relational form, you've separated the issue of what machine the data is stored on from the actual code that you write. Those become completely separate issues. That gives the organization an opportunity to say, This database on DB2 can be on a Teradata machine a year from now. So the move to DB2 is a very positive move for the industry, getting that separation of data from the programming code. It means the data can be put on the most cost-effective platform, without having to totally rewrite all the software.

We offer a product called SQL Server, based on technology from Sybase. So you can have DB2 on the mainframe, with Oracle or Sybase on the minicomputers. And we are the exclusive licenser of Sybase on microprocessor-based systems.

But we're not going to write a database that runs on the IBM mainframe. We are just going to play a very strong role in the standardization of SQL that promotes this independence of data from application.

Q. For the MIS audience trying to decide on the best technology risk, can you clarify what is happening between OS/2 and Windows?

A. It's real simple. OS/2 has not sold well. Microsoft invested heavily in it and we're continuing to invest in it. It's not clear what version will spark its sales. We're committed to it; we're doing new versions.

Meanwhile, what's concrete? Graphical interface is very concrete, and Windows and Windows applications are very concrete. The key message you are hearing about OS/2 is that although it's not all that successful, the key to success is running Windows applications. So whether it's the work IBM is doing in Version 2.0 to run Windows applications, or in Version 3 the way we built the Windows capability right into the system, the safe thing today is that Windows applications are being built feverishly in the industry.

What kernel is underneath is no longer that big of an issue. We've got Windows compatibility in OS/2 and we have great, high-level tools. So people should invest in graphical interfaces and graphical applications now--which means Windows. Where they go in the future depends on how good of a job Microsoft and IBM do with OS/2.

There are people who looked at OS/2 early and feel like they should have waited longer. And there are people who feel they got the benefits by being early. The volumes are very small today.

Q. We see Windows referred to as an operating system and a GUI alternately in the trade press.

A. Windows really most properly should be called an operating system. It has its own scheduler and memory manager. The only thing it uses out of DOS is the file system. We have said with Windows NT that we actually implement Windows without the use of DOS. So it is not even reliant on DOS being underneath. That's something that we have to do a better job communicating, is Windows' role as a complete operating system, and its independence from DOS in different implementations.

Q. Then how does it fit with the network operating systems that we need? There is confusion about that as well.

A. Our message on that is pretty clear as well. We feel that networking capability should be built into the operating system, into Windows. So that when you turn on your system, the separate pieces you have to get today--things like a redirector or remote booting, or being able to browse the network--those things should come built-in.

Q. How did you react to the IBM/Novell alliance, announced in February?

A. Well, everybody in this industry has worked with IBM, no one as much as I have. And it's important for IBM to work with everybody. It would have been nice if they had only worked with us, but hey, they have not had their sales force all that oriented toward PC networking. So not only working with us, but also working with Novell, made sense for them.

They're playing catch up in the networking area. It would have been a huge opportunity for them to be a leader in defining their own product, or having their sales force be the best at it, but they did not seize that opportunity.

Q. Can you comment on the relationship of Microsoft and IBM these days?

A. Well, it's the favorite topic of all daily and weekly publications. The barometer is going up; the barometer is going down. The number of cooperative projects we have with IBM is vast.

Q. Do you put much importance on the Patriot Partners effort?

A. Well, for those of your readers who want more operating systems, this would be an excellent opportunity for them to learn another one and rewrite their applications. At the heart of it is something pretty valid, which is the move to object-oriented systems. That is exciting and legitimate. After networking, the next big theme of development efforts at Microsoft is object orientation.

We're doing that in an evolutionary way, upwardly compatible. With the large investment in software on PCs, you might say that is common sense. But our destiny is to evolve new things in a compatible fashion. It's the destiny of other people to try to come up with incompatible systems and see if they can break into a market at the desktop, which you could call a single standard with DOS/Windows, or two if you include the Macintosh. And we do in our development activities, but a lot of your readers are probably PC-only. I don't see many corporate applications being developed on the Mac side, not even as much as its market share would suggest.

Anyway, we're doing the same kinds of things Patriot is doing, but in our evolutionary fashion.

Q. How quickly do you think mainframe-oriented software companies will respond to the move to client/server?

A. How quickly they will move to this relational, graphical, client/server world is tough to say. They have to bring in new developers to do some things. But there are certainly opportunities to define some application packages that work in this environment, and be very aggressive about that.

The mainframe software companies are really just applications companies. That's really it. Who are the mainframe software companies?

Q. There are systems management tool providers, like Computer Associates and Goal Systems.

A. Yes. But there it's so specific to the operating system, that mostly what you are doing is filling in holes with things that the guy who wrote the operating system forgot to provide. So it's completely different, even the tools that run on VM versus MVS are fairly different.

I don't know if the classic mainframe tool providers will play much of a role. You haven't seen them move down to the PC environment much. And you have a lot of small PC software companies that fill those holes quickly.

Q. Do you think client/server will promote open systems?

A. This word "open" has been defined a lot of ways. Any way you define it, though, the most open system in the world is the PC. People can switch hardware vendors every day of the week. They have more choice in software than for any other platform.

We promoted the idea of strict binary compatibility. The manufacturers were not very enthusiastic about that, but once it got momentum among end users, it carried forward. The more machines that were sold, the more software came out. All the incompatible PCs, with the exception of the Mac, were discontinued.

As you move up the computer chain, you get more and more proprietary. The PC is the most open.

At the midrange, it's not completely open, because you can't take software and move it from machine to machine on workstation-type systems. We're moving up into that arena with our portable versions of Windows, and trying to promote binary compatibility.

But Unix has been very fragmented. The way the Unix vendors used the word "open" anticipated more unification in the Unix market than has come to pass. So unfortunately, the term, which is a very important term, people are almost cynical about because Unix did not come together.

Bill Gates, chairman of Microsoft, has made an indisputable impact on the software industry. Windows 3.0 is now called (by Microsoft) the best-selling product of all time. Gates conducted this interview with Software Magazine Editor John Desmond in June at Microsoft headquarters in Redmond, Wash.