Pacific Connection(英語)

"Open source keeps pounding away"A Conversation with VA Linux's Larry Augustin

In a cubicle at VA Linux Systems, a stream of yellow plastic ribbon---used by American police to cordon off crime scenes--- surrounds the outside walls of a cubicle. The tape is printed with the phrase: "Windows Free Zone," an appropriate sentiment for this Sunnyvale, California company, which has carved out a business integrating Linux with Intel-based machines that it sells as a server solution---an alternative both to Windows NT and Sun Solaris. Along with Red Hat and a few other companies, VA Linux stands at the center of a movement that has transformed Linux from an underground into a commercially viable server operating system.

The "VA" in VA Linux stands for Vera and Augustin---that's James Vera and Larry Augustin, who as Stanford students, started VA. Vera is no longer active in the company, but Augustin has made VA Linux almost synonymous with the operating system---at least as it applies to commercial development. In the process, he has become a multi-millionaire---at least on paper---even with a subsequent dramatic slide in VA's stock price.

At first glance, Augustin hardly looks like the kind of guy to be running a Linux company. After a stint at Bell Laboratories, where he worked on high-speed digital switching, Augustin attended Stanford University where he pursued a Ph.D. in electronic automated design tools---the software used design new microprocessors. His main focus was a verification prover---a tool that verifies that a circuit is behaving in a desired way. In retrospect, Augustin freely admits that was a pretty obscure project for a guy running a Linux company.

But Augustin was also a self-described computer hacker. He played with hardware, wrote software, and used just about every operating system ever written. The turning point came when Augustin wanted his very own Unix machine. While he admired Sun, the $7000 price tag of a Sun workstation was out of the question. Instead, he build a PC running a two-year-old operating system called Unix. The resulting machine ran faster, yet cost less: about $2000. Other people took notice, asked Augustin to "build one for them," and a business was born.

From its founding, VA Linux has had a business model closely based on two companies: Sun and Dell. Sun---because of its maverick management and overriding success in the Unix server market. Dell---because of its stature as the best-known place on the Web to order a computer. Not surprisingly, Sun is now a competitor. But lately, so is Dell, which has added Linux servers to its online catalog. IBM too has entered the market, recently agreeing to bundle Red Hat's Linux release with its own servers.

With so many sharks in the commercial Linux waters, how does a much smaller company like VA Linux survive? VA has a one-word answer: culture. The company is mindful that Linux development is conducted not by a single company and its programming staff, but by a community of loosely-coupled developers: some working out of self-interest, and a shrinking number driven more by idealism. Augustin and company argue that if you nurture the developer community--- staying close to Linux's cultural roots-the community will return the favor and your customers will see the difference.

And so last January, VA inaugurated SourceForge, an online breeding ground for Linux projects that currently supports nearly 30,000 registered users and hosts some 4,500 open source projects. Or point your browser to, and you will arrive at a "Linux Community Production," quietly staffed by VA Linux personnel but pledged to corporate neutrality. Last June, VA acquired Andover.Net, and its two developer-savvy websites: ("News for nerds") and, which claims "the largest index of Linux software on the Web." (All VA Linux websites converge at VA's newly announced Open Source Development Network: OSDN.Net.)

It remains to be seen whether this exemplary citizenship---and the developer contacts it provides---will enable VA Linux to beat off the competition. But one thing is for sure: no other Linux company has "Linux" in its name. If name-brand appeal---and a core belief in the open systems development model---means something to customers, VA has got it made. On the other hand, if customers merely see Linux servers as interchangeable boxes, VA has its work cut out.

And so Augustin is taking nothing for granted. When he's in town, you can often find him hard at work at his cubicle (at this writing, he eschews a walled-off office), whose decorative style include a stuffed penguin---the Linux mascot---but no yellow police tape. Off hours, such as there ares, Augustin occasionally sees Linux god Linus Torvalds: their kids play together. I caught up with him by cell phone.

Let's start with the basics. Who are Linux's most promising customers?
Right now, Linux is more frequently purchased by technically savvy companies than by companies for whom technology isn't as core to the business. If you are a dot-com, you are likely to buy Linux. If you're [a large department store chain] you are just not as likely to think about Linux.
Is that because Linux still requires a leap of faith?
It's not so much a leap of faith as it is mindshare and history. Traditional companies that are not dot-coms prefer technology that has been around for five to ten years. They don't have a high degree of tolerance for anything that's borderline new. Even the whole concept of the Web is new to them.
Open source doesn't have an official development "staff". Does that still make people nervous?
It does some, but not nearly as much as it used to because there are plenty of companies taking on that role. One of the great myths of open source is that there's no one to support it. But we don't see that as an excuse when we go calling on companies. VA can give the customer a roadmap that is at least as accurate as the one Microsoft gives.
It's a different model.
Yes, it's a different software model--- which does take some understanding to get over. We tell people that because Linux is open source, you have advantages. For example, if you want to leave us as a customer, you can. You're not locked in. Of course we're going to hold your hand the whole way and be invaluable to you so that leaving us will be difficult. But fundamentally, if you leave us, the code is still there, somebody else could pick it up and support it. You're not left out in the cold if we go away or if we don't support you properly. Whereas if you get tied into Microsoft, what are you going to do? For us, the millions of people out on the Internet working with the code are a selling point.
Does it bother people that Linux developers---as opposed to Microsoft developers---aren't necessarily motivated by profit?
Yes, it does. People constantly ask whether the community going to continue to develop the software: and why. They constantly ask that. The answer is that a lot of the developers are employees of IT organizations. There are a lot of system administrators that make up the Linux developer community. Why are these people contributing to Linux? Because the organization they work for uses some piece of free software. The classic example is the Cisco Enterprise Print System (CEPS), which is a basis for a lot of the next-generation printing work in open source. It was developed within the Cisco IT organization because they needed it. They went open source because printing isn't a core piece of their business and they weren't necessarily concerned about having intellectual property around printing. They just wanted to print better. But even more than that, they didn't want to be the only user. They wanted to make sure somebody else cared about what they were working on. They wanted there to be other users. That's part of the reason they went open source and started giving the code away. They wanted other people to be dependent on this technology as well.
You're saying that this is not an act of charity, but of self-preservation.
Yes, this was a deliberate strategy within the IT organization to create a piece of software that they needed for the infrastructure---so let's get other people to need it, and contribute to it, as well. So let's think about the economics of this model. The alternative is to go to a commercial company and buy a piece of commercial software. That company develops the software, then goes out and sells it. The money they make from the sales goes back into R&D, but also into profit, overhead, sales and marketing. And so it turns out to be more economically efficient for IT organizations to collaborate on a common piece of software, than to go out and buy that piece of software from a commercial developer.I think this is the big economic driver behind a lot of open source. We had another example within VA. We needed a customer relationship management [CRM] software solution. We could have gone off and spent multiple millions of dollars to get it. Instead, we spent $150,000 on an open source project. We funded a data project in part by contracting with some people outside the company, but we also hired a couple of developers inside VA to work on it as well. Again, it turns out to be more economically efficient for each IT organization to contribute two developers than for each of us to contribute a million dollars to a company.
But projects like these are only feasible if you find other people to share the risk, and are willing to contribute with you. Open source makes that possible. Did you put that application back into the public domain?
Yes, and now other people are growing it. In fact, one result is the emergence of a new company, called Open CRM, that is part of the group of people that we funded and worked with. They are going to sell service and support for Open CRM, and will do customer implementations.
You've put a lot of work into SourceForge---your collaboration website. Is that simply a charitable act on your part to foster the community? What's in it for you?
We believe our business depends on the growth of open source and the open source community. We want to see lots of open source options to feed into solutions for the customer. By creating SourceForge, we help to grow that library of options we have at our disposal. The more open source software, the more open source developers---resulting in more things to sell customers and more markets for us to enter.And so in some sense, the answer is yes, we did it for the community. But the point is that we've built a business that benefits from growing and developing open source. We've structured the business that way, and that's a key thing. It means we're removing incentives to develop proprietary software within VA.
What about It's a great source of information, but it doesn't seem tied into Linux.
No, it's not directly Linux-based. But you find that a lot of the people who live and work on Slashdot are open source advocates or developers. We consider Slashdot the discussion community part of the effort--- a place people can hang out and just talk about things. Discussion is an important piece of the puzzle. I think we can leverage Slashdot into a hierarchy of discussion forums around individual projects, around issues that matter to people around Linux and open source.
Wall Street's take on Linux seems like a love/hate relationship. Has the American investment community figured Linux and the open source movement out?
Wall Street recognizes that there's something going on, but they don't understand who is going to make money and how. Wall Street looks around and says, "yup, there's something happening here---we know Linux is taking over, but we still don't understand how Linux companies are going to turn free software into revenues." That's where the love/hate relationship comes from. They love the idea that Linux is taking over by winning market share. But they hate it because they don't understand the business model.
Some analysts acknowledge that Linux sales volumes are high, but there's still concerned about revenues. Is that something you have to keep explaining?
One of the things we have to keep explaining is that we're not Red Hat. Yesterday, for example, [analyst] Phil Rueppel of Alex Brown upgraded our stock. But even so, he maintains that investors remain skeptical about the ability to generate revenues in the Linux market. A recent report from IDC shows that Linux revenues will grow to only $80 million by 2003. That last statement immediately says to everyone, maybe they've been upgraded but there's no chance for revenues in this business --- why is anyone even in the business?
What's wrong with that statement?
That's revenues of Linux CD sales [Red Hat's core business]---completely unrelated to what we do as a business. If you go back six years ago to our first slide presentation and business plan, we've clearly modeled ourselves on Sun and Dell, right from the beginning. That was always my plan because I never felt that CD sales was the way you could generate a big business..
Is it more a Sun model or a Dell model?
There are parallels to Sun, but at the same time, we do build-to-order on the hardware. You can go in and select the hardware configuration. We maintain zero inventory. We go off the shelf as much as possible for the PC volume market.
Do you build your own boxes?
Yes. But we don't build ASICs, for example, or other chips. We get other people to build the motherboards. And we don't own the exclusive rights to those boards, so we can leverage off of additional volume. Those are all the things that make our model a Dell-like hybrid..
The topic of Linux on the client won't go away. Is that an area you want to get into?
One of the interesting things about open source is that it doesn't go away. Once it starts pounding on something, it may take years to get traction, but it never goes away. Linux on the client is like that. I think it's inevitable just because there will be a constant barrage of client side software appearing for Linux. It may take a long time because there's an incredible entrenched community or marketplace around Windows. It could take 10 years. At VA, we want to be in a position to take advantage of that as it happens and help it out along the way. But from a business point of view, it doesn't make sense for us to beat our heads against that market right now..
What about the open source database MySQL versus the widely established commercial database Oracle?
In a sense, that's the same question. Today Oracle has the features, power, and support. MySQL is not there. Technically Oracle is superior. It offers real transactions and performance. MySQL doesn't offer that, but it has advantages in the entry level. It's leaner, faster, easier to use, has an incredible number of interfaces. You find that MySQL runs all of the smaller websites and Oracle runs all of the much larger, more mission critical pieces. .
Is open source at a disadvantage on high-end machines because there are fewer high-end platforms, and therefore fewer programmers to develop it?
It takes open source longer to get to the high end because there are fewer developers with access to the high end environments. But remember: most IT developers are being paid by companies. As open source becomes more accepted, you find more and more companies hiring developers to support open source on larger platforms. People will develop less as a hobby and more because it's a part of their job..
So you would make the same case with other Linux development issues. Principal development will eventually be driven by professional programmers working for their company's self-interest.
That's right. In fact, in open source, technical superiority is not necessarily the best indicator of what's eventually going to succeed. The strength of the developer community may be more important than current technical leadership. Let's go back eight years when the contest was between Linux and BSD. I'd download off of the web open source 386BSD from Bill Jolitz [the original developer of 386BSD] and clearly, it was technically superior to Linux. It had killer networking stuff that far outperformed Linux; it had features like a networking stack. It had all these things. But Linux had a larger, more vibrant, more active developer community.I think if you do a study, you'll find that while a particular solution may be technically better today, if it doesn't have the mindshare of the developer community, then it's probably not the solution of the future..
What was it about Linux that attracted that community?
The development model. The BSD development community was very cathedral style [a reference to the Eric Raymond essay comparing "cathedral" and "bazaar" methods of software development.] It was very closed. It was a small clique. They tended to only let in people who had really proven themselves. For example, Bill the would say "okay the next release is going to be in nine months, and he'd go off, people wouldn't hear from him for nine months, and then a new release would come out. By contrast, Linus [Torvalds] would put his set of files across the server every night. People would send him patches, and he'd say okay, here's my new code..
Did Linus understand what the early Internet represented in terms of community development?
I don't think he understood it consciously. There was no grand plan. He was just writing code, and started doing what seemed to work. I remember the first piece of software I put out that was under the GNU GPL [General Public License]. I had done a version of Bison, the GNU parser generator, for C++, and I remember getting back patches that were over my head---they did things I didn't know you could do in C++. I'd put the patch in; everyone thought that was cool and that I must be smart. I'd say "no, someone else did it for me."
So I don't know that there was any grand plan. It's just that Linus did what seemed natural and seemed to work. By contrast, the BSD people didn't really have this notion of opening the process up..
Is VA Linux active in Japan?
Yes. Sumitomo, our lead partner in Japan, announced back at LinuxWorld Japan that we are launching VA Linux Japan in conjunction with VA and some other partners..
How different is the Japanese market?
I don't see it as all that different. We're not the experts in it, of course, which is why we're bringing in several Japanese partners. But from what I've seen at LinuxWorlds in Japan over the last two years, there's an incredibly strong Linux developer community there, just as there is worldwide. The Japanese "speak" Linux and open source in the same way we speak Linux and open source. I recall at the show sitting in a group of 20 Linux open source developers. Even with the English/Japanese language barrier, we could still understand each other well enough to talk about Linux and open source and what projects people are working on, where we see projects going and what needs to be done. You don't need to speak fluent Japanese to see the talent within that community..
Do you think there will be the same corporate hesitancy there as here?
I think there will be less in Japan. For one thing, Linux is not a U.S. product. I think that's a big advantage outside the U.S. When you go into Japan, you want support from people like NTT, NEC, and Toshiba. And with Linux, they are not beholden to a U.S. company. Sure they can work with VA, but it's not the same as being dependent on Microsoft. That principle applies worldwide..
If Linux had had a head start on Microsoft, would there still be a Microsoft as we know it? Is Linux's come-from-behind position simply a matter of timing.
That's an interesting question, and it just might be the case. On the other hand, the environmental conditions had to be right to create open source and Linux, just as the conditions had to be right to create Microsoft. You might say that it wouldn't be possible to have the conditions for Linux unless you first had gone through the conditions that could create Microsoft..
I have to ask you about your infamous cubicle.
I don't like offices---but I think I may be compelled to have offices at some point. There's a great book by two famous authors who have done software engineering studies. They found that something like five interruptions a day can completely destroy the productivity of someone doing engineering. They also found that every time your phone rings, you lose about fifteen minutes. And so every interruption is a tremendous time drain. For that reason, we may end up having some offices.
But given that the company president is usually entitled to an office, why a cubicle?
I'm no better than anybody else. I just happen to have the job where I coordinate everything. Is that more or less important than the guy who is writing some important code? I struggle with that concept. I have higher walls on my cube, which is as far as I was willing to go with it. I also think an open office leads to more cooperation-an important part of the environment that you can foster better without walls..