Open Source Evolution

What is the next stage in the evolution of open source? Companies like Red Hat, VA Linux, and Linuxcare are betting they can build businesses without selling proprietary software, but can they?

Opening graphic

We are well into the age of astronomical investments and driving
valuations of Internet companies. Eyebrows are no longer raised
when a risky start-up with an unproven track record scores multimillion-dollar investment from venture capitalists. Most analysts and market watchers agree that not ALL of these companies can succeed, but time and again they predict that some of them will. The process, then, is one of laying bets, and investors are betting heavily. Indeed, Linux and open source ventures are lately among the most popular draws for this kind of speculation.

Efforts have been made by many of Linux’s staunchest supporters to convince the community that Linux (and indeed all software) isn’t really a product, but rather a service, and that the fact of this distinction will ultimately make untenable the proprietary business model for software. Open source advocates have argued this point convincingly; but however compelling these arguments may be, the fact remains that only one software business model has proved capable of driving an entire industry and making millionaires of geeks the world over: Microsoft’s.

So, with the Y2K doomsayers temporarily unemployed, pundits are now turning to the Internet economy, and the open source companies that have spun out of it, looking for signs of collapse. Though he doesn’t fall into the “doomsayer” category, Larry McVoy, CEO of Bitmover, Inc. and a former top scientist at Sun Microsystems, says it remains to be seen whether the business models of some of the most successful Linux vendors will be viable in the long term. Open source business models are further complicated by a unique factor: developers of free software are not always in it for the money, and idealism rarely drives profits. Without appropriate margins, argues McVoy, Linux companies cannot do for free software what commercial entities do best: funnel money into new projects.

McVoy says that open source advocates like Bruce Perens, Eric Raymond, and Richard Stallman can only see the benefit of open source software from the end-users’ perspective. “Their arguments are somewhat shortsighted,” McVoy insists, “because they’re fighting against the sort of evil world where it’s been all proprietary and you couldn’t get the source code. I’ll be the first to admit the proprietary people have abused their end of the spectrum. But in solving that problem they’re creating a second problem. They’ve gone so far in the other direction, they’re making it very difficult for people to actually generate revenue off of stuff in this new space. If you can’t generate revenue, you can’t create new stuff.”

The Price of Innovation

McVoy believes that, in the long run, pure open source companies will not have enough margin to continue to fund innovative software development. He also says that, ironically, open source companies are cannibalizing the sales of large corporations like HP, IBM, and Sun, whose revenues paid many of the engineers that contributed much of the open source code being used in Linux today.

Take BIND (the Berkeley Internet Name Daemon), for example. For years, proprietary software vendors funded the Internet Software Consortium to continue development on this standard piece of Internet software (BIND is the software that translates URLs like www.linuxmagazine.com into IP address-es) — software that companies like Red Hat now include as part of their standard Linux distribution.

Do open source companies really innovate? “[Richard] Stallman will tell you that you can create revenue. ‘Look at [development tools company] Cygnus, they generated revenue off of support,’ he will say,” says McVoy. “But what did Cygnus ever do that is new? The only time they did anything new in their history is when they went and got VC funding. They were never able to do anything significant that was new based on the support revenues, nor has any other company been able to do anything significant based on support. There just isn’t enough margin there.”

Michael Cowpland, CEO of Corel — a company that built its business on proprietary software, but is betting its future on open source — disagrees that the margins are so slim: “A major portion of revenues in the computer industry are tied up with know-how and support and service,” says Cowpland. “The amount embedded in the actual proprietary IP is much less than people originally thought. You gain more than you lose by being on top of the open platform. Each company can then specialize in what it does best.”

Competing the Open Source Way

Certainly, major open source players like Red Hat, VA Linux Systems, and Linuxcare are banking on being able to foster innovation while at the same time avoiding the sale of proprietary software. The employee rosters for these companies read like a who’s who of Linux development.

But in order to succeed, they will need to give their customers a powerful reason not to jump ship when a competitor walks in and offers a similar product at a lower price. The need to differentiate helps explain why a company like VA Linux would spend close to $1 billion to acquire Andover.net and its well-branded Slashdot and Freshmeat Web sites. Since its acquisition of the Linux. com domain in 1999, VA has tried to leverage its close relationship with the Linux community as a key point of differentiation from other hardware vendors. Without strong brands and differentiation, a company like VA could be in serious trouble if a much larger player, say Dell Computer, decided to go after its core Linux hardware business with full force.

Chris Peterson, an analyst with Palo Alto, CA technology think tank the Foresight Institute, does not believe that this kind of differentiation must be derived from proprietary software. “The customer is not paying for the source code,” she notes, “so the actual services [provided by open source companies] are not that different from a proprietary company.” In fact, she believes that pure-play open source companies have the opportunity to make significant profit margins. “One of the main effects of open source is becoming a standard,” she says. “The profit potential is huge once you’re a standard.”

While proprietary vendors have bankrolled much open source development, or at least paid the salaries of many key open source developers (see timeline below), the idea that proprietary companies are necessary for open source innovation is a touchy subject.

timeline, part 2
timeline, part 1
timeline key
A LOVE-HATE RELATIONSHIP: Commercial vendors have contributed to many important open source projects, but not always with stellar results.

“Even before we had money from investors, Red Hat was actively participating in the GNOME project, which was new and innovative technology,” says Red Hat Enterprise Business Unit General Manager Paul McNamara. “We [fund innovation] based on revenue generated from the sale of our products and services. We recognized that tech leadership is an important part of our business plan so it has to be a component of our spending.”

And a look at the genesis of many key open source projects reveals that many of them — Sendmail, the Mosaic Web browser, and Emacs, for example — had their genesis in the academic, rather than proprietary-soft-ware, environment. From Sun Microsystems to Netscape to Lucid, Inc. there is a long list of vendors who have dabbled with open source software projects with varying degrees of success.

But to the investors in today’s high-flying open source companies, the question of where open source technology came from is moot. To them, one question rises above all others: How do you make money on open source software?

McVoy may not claim to know the answer to this question, but he has some interesting ideas.

Hybrids: Breeding Success

Evolution Young

The moment we sell proprietary software we limit our opportunities, because we will have given up on the bigger opportunities available to us.”

– Bob Young, Red Hat, Inc.

“In general, whenever you have big margins, you either have to have some lock or people will take your business away,” says McVoy. “The real flaw in the whole open source model is there isn’t any lock. As a matter fact, the OSD [Open Source Definition], from a business point of view, is designed to make sure there is never a lock.”

A number of companies have already implemented hybrid open source strategies, or designed ways to compensate open source programmers who had been contributing code for free. The giant hardware and software vendors are all toying in one form or another with the hybrid model, selling products, services, or support for free operating systems or using them to funnel buyers to lucrative hardware sales.

Caldera is one such company. The Utah-based company releases its distribution of Linux for free and under an open source license, but adds its proprietary Lizard installer to the package. Similarly, TurboLinux adds proprietary clustering technology to its distribution of Linux. Corel is likewise mixing proprietary and open software. It expects a strong for-profit business from its applications software for Linux. Corel’s Office 2000 for Linux went into beta testing in late January and is slated to ship to customers in March.

The Foresight Institute’s Peterson, however, has more confidence in pure open source plays. According to her, the idea of a product space and license space for hybrid companies has not been explored as much as it needs to be. “There will be a lot of new things,” she observes. “I don’t have anything against hybrid plays; there’s exciting stuff that can be done there. We personally think pure open source play is the way to go if you want to be the new standard.”

Though he will not rule out ever selling proprietary software, Red Hat Chairman Bob Young sees the adoption of a hybrid business model as an opportunity lost. “The moment we sell proprietary software,” he says, “we limit our opportunities, because we will have given up on the bigger opportunities available to us.”

McVoy believes that partial proprietary approach is indeed the best solution. “I think there is actually some value in having some ownership of the software,” he said. “I don’t want to see the world return to this proprietary closed source model, but I do want to see some model in the open source world where people can actually generate some margins. I want to see Red Hat have enough people and time and talent and money that they can solve the clustering problem and other problems that take a lot of effort to solve.”

Bitmover: A Hybrid Case Study

Evolution McVoy

Open source advocates are making it very difficult for
people to actually generate revenue off of stuff in this new space. If you can’t generate revenue, you can’t create new stuff.

– Larry McVoy, Bitmover

McVoy’s hybrid model for his own company, Bitmover, was spawned when Linux developer Alan Cox came up with the idea of a central log for changes to the Linux kernel. McVoy decided to develop a tool to manage this, and Bitmover, Inc.’s source code management product, Bitkeeper, was born. Bitkeeper is not yet finished, but it is being used by the Linux PowerPC development team.

With the hybrid license that McVoy developed for Bitkeeper, he “was trying to make sure all…parties involved won. The open source definition is a model where the people who [win] want open source, and they’re the only ones that win. The people who wrote the code do not win. The people who don’t give a damn if it’s open and just want to be supported, don’t win.”

For the open source community, Bitkeeper was designed to be free and open so long as developers agreed to publicly log their code changes and bug fixes into the system. For commercial organizations that valued privacy and were not willing to publish the logs of their code changes, Bitmover would charge for the product, and it would remain proprietary. McVoy created a hybrid open source license to accommodate this.

“For the people in the open source world, we have a Bitkeeper license that says when you make changes, you log what you did. The ‘this is what I did’ comments get publicly logged. That’s mandatory if you’re using the stuff for free,” McVoy said.

For those who prefer to see the world in green and white, the list price of the private Bitkeeper license is $2,600 a seat per year.

“Why wouldn’t I just use it for free and put up with the logging? The reason is that those logging messages include the e-mail messages of the user who made the changes. A happy hunting ground for headhunters!”

McVoy says he’s had long arguments about this with Richard Stallman and Eric Raymond. He says that by forcing the change logs into the open, he is actually doing the open source community a service.

“The community wants those change logs. Right now if I do internal modifications to something that’s GPLed or any other open source license, there’s no requirement, unless I ship that stuff to some other person, that I ever give any of that back. There’s all sorts of work going on that you don’t know is happening.”

McVoy is impatient with the attitudes of some open source advocates, whom he describes as license police. “This is a self-elected class of people — Perens and Raymond and the OSI board and so on,” McVoy said. “They decided they represent the community without any sort of elective process at all. They decided to police what licenses they like and what they don’t like. They hate the Bitkeeper license with a passion.”

For his part, Raymond says he simply does not consider the Bitkeeper license to be Open Source. “The license itself does not anger us; Larry has a perfect right to construct any kind of license for his creations that he likes,” says Raymond. “Larry’s attempts to pass it off as Open Source by arguing with us about what the Open Source Definition says and what it means have annoyed the Open Source Initiative in the past.”

Well, perhaps McVoy will have the last laugh or, at the very least, be able to pay someone a substantial amount of money to have it for him.

The Sun That Might Have Been

Larry McVoy’s belief in open source is unquestionable, going back to 1993, when he wrote a paper that he presented to Sun Microsystems Chairman and CEO Scott McNealy. McVoy was convinced that Sun was all wrong in its operating-system strategy. Sun was in the process of converting from Sun OS to Solaris. McVoy hated Solaris.

At this point in Sun’s history, the company was worried about having to pay per-seat royalties because its operating system was derived from AT&T Unix. Every time it made a sale, Sun had to send AT&T a check. McVoy started hacking around inside Sun OS and removed any code that was AT&T specific, similar to the manner in which Netscape later removed any proprietary code from its browser software so that it could open it up to the industry.

“There wasn’t actually that much in there, and I made the system boot and demonstrated that it would run almost anything any other operating system would run,” McVoy says. The rest of the code was free, derived from open source BSD code and other free software.

In a paper entitled “The Sourceware Operating System Proposal,” McVoy wrote that “software that is widely available and royalty free is more useful and valuable to the end user than proprietary software.” (This would later become a mantra for Red Hat and other companies. The McVoy paper, in fact, was a key influence in the creation of Red Hat’s business plan back in 1993.)

In his proposal, he made a business case for open source software. He didn’t argue that it would make Unix better, but rather that it would solve all the fragmentation issues of Unix and give it a fighting chance against Windows NT, whose capacity to deliver a single specification seemed to pose a serious threat to Unix.

Sun never adopted his plan.

Wendy Goldman Rohm is the co-author of Under the Radar, a book about Red Hat. She can be reached at wendy@ linux-mag.com.

Comments are closed.