CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

Singing hosannas for Linux

Big Blue's Dan Frye on why Linux is key to IBM's future and how it will sink rival Sun Microsystems.

At first blush, somebody with a background as a theoretical atomic physicist would not jump to mind as the obvious candidate to take on the task of figuring out what IBM should do about Linux and open source.

Dan Frye quickly acknowledges as much, adding that he is not "an OS guy" by training or temperament. Yet the luck of the draw three years ago elevated Frye into the position of shouldering responsibility for proselytizing on behalf of Linux within the ranks of Big Blue.

The co-author of IBM's first strategy papers on both Linux and open source made a persuasive case, ultimately convincing CEO Lou Gerstner to put this on the company's strategy agenda. Indeed, Gerstner last November announced IBM's intention to invest $1 billion on Linux in 2001.

The world spends way too much time and energy doing exactly the same thing in a proprietary fashion. That's just wasteful. Nobody gains by that. As director of IBM's Linux Technology Center in Beaverton, Ore., Frye, who oversees the activities of approximately 200 people, will play a large role in furthering the company's ambitions to push Linux into corporate America. He recently sat down with CNET to discuss his thoughts about the future of Linux and its potential to shake up the constellation of power in the operating systems market.

Why did IBM establish a Linux Technology Center?
Our job really is to help make Linux better.

How about the specifics of how you get from here to there? After all, you're not working with an internal development team but with a large number of people who don't report up through the ranks at IBM.
The way to do that is not by telling the Linux community to "Do X." You say, "We're going to do X." Then you deliver it and wait for the community to accept it. So, for instance, we have lots of people on things like scalability or file systems or interoperability, libraries, etc. There are a number of things in Linux that IBM would like to see made better.

How long have you been involved in promoting Linux to IBM's brass?
I started working with Linux and open source more than two and a half years ago as part of IBM's emerging technology group. I had been there just a week and during our regular meeting asked what else should we pay attention to. I got the assignment to go ferret out what to do about Linux. I came back a week later with a one-page summary and gave to the VP in charge of strategy.

What was the reaction?
She said, "You've got to do something about this." In upper management, there was no resistance whatsoever.

What was the argument you made for Linux vs. the arguments others made for teaming more tightly with a proprietary OS?
The value attributes were so obvious. Because Linux is the same everywhere, if you're writing a brand-new business application and want to run on Unix, then Linux becomes the obvious choice. Once you write on Linux, you're not beholden to a particular architecture or software vendor. If you don't like the hardware, you can change it. And you don't have to rewrite your apps. It's the value proposition that Unix tried to do 15 years ago but failed.

Linus is not a gate; Linus is a control point.Lou Gerstner gave a speech in November, where he talked about IBM's plan to spend 1 billion on Linux this year. How difficult a sell was it to convince the folks running the corporation that this was a good thing to invest in?
The open source sell to the senior executives at IBM was extremely easy. That wasn't necessarily the same thing with middle management. Sometimes that took a little more time and effort. But the senior executives just picked up on it. We dispelled a lot of the myths about open source. That was the big thing--the myth that open source was not secure or the myth that it was undisciplined. It's highly disciplined--more disciplined than a lot of proprietary shops.

Any risk that history might repeat itself--but with Linux becoming a victim of the same kind of intra-community competition?
With the open source model, you really can't screw it up. Because its GPL, Linux is not going to fragment. Now we don't see this as a competition with Windows. In ten years, we know there's going to be at least three operating systems. There may be more, but there are going to be at least three: OS 390, NT and Linux.

What about time to market concerns as far as developing for corporations? The Linux kernal approval process has to go through Linus Torvalds. Does that vetting procedure add extra time into the process?
It actually makes for a faster process. It becomes a slightly unpredictable one, but still it's a faster process. Linux is very modular as one of the design principles that you add things and you don't need to change the entire operating system. So, if an IBM wants a new feature in Linux, a file system or a new volume management system, you would do it according to the rules of how the community works. You show that it works, put it up on the Web site, and get people to use it and feel comfortable, etc. At some point, when it's good enough, it'll go into the kernal.

So you still see Linus' active participation as a net-net positive in expanding the functionality of the operating system?
If you do good work, it will be accepted. Linus is not a gate; Linus is a control point. But we're not worried about getting the functionality. If you ask the Linux community to do something, they may or may not get around to it. But if you're willing to participate and write code in such a way that it's good enough for them, it'll go in.

What's your pitch when you talk with customers in corporate America about Linux?
In terms of corporate America, Linux represents a different philosophy. But customers don't really care about the development process. What they really care about is whether their applications will work on it. Frankly, though the development process from the outside looks a little bit chaotic, it's highly disciplined and moves faster because Linux--and most open-source products--is very modular. As I've said, you can have many more people working on it without it affecting the rest of the code.

From an IBM perspective, is the idea to play up the cost factor when you sell corporate customers on Linux?
Actually, cost is not what makes the sale for Linux. It's the portability of the operating system, the fact that you can build your business on Linux and can change hardware vendors or architectures. For corporate customers, you need to provide a highly reliable platform, and they prefer to have a choice. So with Linux, they have a choice of vendors or a choice of architectures, and they're not buying into a single corporation or a single solution. Their technical staffs love the flexibility and the fact that it's open source; they can see what's going on.

What's IBM's thinking about open source? Microsoft's Jim Allchin recently had some harsh words, suggesting it stifles innovation. Do you see agree?
No. Open source is good for business. Now I should add that open source is not for everything in software. We have a very large and successful software business, and we're going to retain that. But open source is great for infrastructure code. The reason is that to make open source work, there has to be an overlap between the people who care about the software and the people who make the software better. As you get further up the application stack, those two groups become disjointed...so the software that checks you into a hospital will never be open source because the people who care about that can't write software.

The merits of one way of writing software over another--it's kicked up a lot of dust. Are you surprised?
The world spends way too much time and energy doing exactly the same thing in a proprietary fashion. That's just wasteful. Nobody gains by that. With Linux, you get people who work together and share innovation. The software is better and we, along with everybody else, get to spend extra programmer hours writing software that you can make money on.

Where does Linux need to go to become more widespread in the enterprise?
Most of the customers use Linux on the edge of the enterprise--in Web serving, file serving, etc. You're beginning to see a fair amount of use closer to the center of the enterprise. For instance, we have dozens of customers in trial evaluations of Linux on the mainframe who are spending lots of time and effort trying to see if Linux is ready for them. By and large, they haven't started buying yet. But we expect to see some significant movement there.

Is it a technology issue or one of perception?
It's both. More people have to get comfortable with Linux. Also, things like scalabilty and serviceability, threading file systems, security, documentation--these are all important. There are thousands of apps for Linux, but an enterprise OS has tens of thousands. I think Linux is really good enough today for a lot more things than it's being used for. Is it ready for a mission-critical, center of the enterprise today? No, but it'll get there.

What's the challenge then?
The Linux community needs to figure out a way to work together better to enable more apps faster. If there's anything that will cause Linux not to succeed, it's because we don't figure out a way to get enough applications and enough ISVs to port to Linux. That's the hard problem because that typically takes a long time.

How do you think the emergence of Linux will affect Sun Microsystems?
Sun will have problems. They've got one answer to every question: Solaris--whether on the desktop or the workstation. Solaris ties people to the hardware though, because it's a proprietary solution. Sun will respond in one of two ways: They can try to become a heterogeneous company, which will be hard to do, or they can ignore it altogether, which is the path they have been on.

How much longer do you think Sun will choose, as you suggest, to ignore Linux?
Well I hope they ignore it for a long time. I'm just delighted with every day that goes by.