Open War

It started as a crusade for free source code. Linux zealots turned it into a full-frontal assault on Microsoft. Now the battle for the desktop could snatch defeat from the jaws of moral victory.

By Russ Mitchell

Wired

October 2001

Here's one way to coax desktop computer users away from Microsoft Windows and over to Linux instead: Wait until they're not looking, strip all Microsoft programs off their hard drives, and leave them to ponder the error of their ways. Lost data? Too bad, baby. The open source revolution has arrived, and you're part of the problem. Deal with it.

That's exactly what happened to Anne Speedie, then an editor at Wide Open News, a Web site owned by Red Hat, the best known of the 140 or so distributors of the open source Linux operating system. Speedie needed to use Microsoft Word because the Linux word processors at her disposal were saddled with spellcheckers so abysmal they caused more problems than they solved, skipping over misspelled words and offering bizarre alternatives for words spelled correctly. One program, Applixware, stumbled on Web site. Fair enough - the term can be spelled several ways. But Applixware offered none of them. Instead, the program suggested podesta. Yes, podesta. Look it up.

Clearly, Speedie couldn't depend on such software and expect to maintain a high-quality Web site. While Red Hat management had asked employees to use Linux programs when possible, it did flash the green light on any software essential to getting the job done. But this didn't sit well with computer support, or with many of Red Hat's software engineers, who, like much of the Linux community, share a monomaniacal antipathy toward all things Microsoft. Speedie, preparing for a business trip, turned her laptop over to a technician to get some dialup software installed. When he returned the machine, Windows was gone - along with all Microsoft applications and Speedie's work files. Outraged, she complained, and when her boss (that would be me) confronted the technician - a stringy little guy in a black trailer-trash T-shirt - he simply stared back and smiled.

After 20 years as a business and technology journalist, I joined Red Hat for nearly a year at the height of the tech stock craze. I was editor in chief of Wide Open News. Tough economics led to layoffs and I got whacked; I left with three months' severance and enough stock options to make a down payment on a house in San Francisco. Nothing to be unhappy about. I also left with great respect for the Linux OS, which turned out to be as stable, flexible, portable, powerful, tight, and efficient as its supporters claim it to be. Technically speaking.

That's pretty much how I feel about Linus Torvalds' kernel development group as well. (See "Keepers of the Kernel [ http://www.wired.com/wired/archive/9.10/kernel.html ]," next page.) When it comes to business savvy, on the other hand, the Linux community is a muddled and unfocused lot. A prime example: the bound-to-fail attempt to compete against Microsoft in desktop computing. Like the Japanese soldiers on Guadalcanal who refused to surrender years after the bomb ended World War II, Linux zealots remain obsessed with beating Microsoft in desktop computing. Desktop computing? Don't they know? The war is over. Microsoft has won.

At best, this obsession is a distraction, a waste of time. At worst, it's sapping energy from the Linux movement and keeping the OS from gaining traction in a market where it actually stands a chance - as a server operating system in enterprise computing. Consider: Linux grabbed a 27 percent share of server OS shipments in 2000, according to IDC - sales that otherwise would have gone to Unix vendors and Microsoft. Linux boosters insist that if free downloads and pass-arounds were counted, that figure would be even higher; and they're probably right. Linux is expected to do as well or better in 2001. This is an area that has real potential.

Conversely, Linux managed only 1.5 percent of shipments in the desktop market in 2000. And that sliver is unlikely to grow in 2001. PC makers are concluding that consumer Linux is too small a market to mess with: Dell Computer recently dropped Linux from its desktops and notebooks.

Given these numbers, it's clear the desktop is receiving disproportionate attention from the Linux community, in the form of anti-Microsoft ranting, misplaced evangelism, a waste of mental energy, and - most important - the draining of developer hours. Legions of programmers are working to break Microsoft's iron grip on the desktop. There's an open source desktop interface project called GNOME, for GNU Network Object Model Environment. And one called KDE (K Desktop Environment). Developers are writing drivers for PC devices whose manufacturers don't see enough of a market to provide Linux drivers on their own. Open sourcers are duplicating work to produce office suites - Sun's StarOffice, AnywareOffice (formerly Applixware), KDE's Koffice, and more.

All this work, and ... it doesn't matter. Desktop computer users care about what they can do on their machines. They want reliability, simplicity, access to popular software, and the ability to communicate easily with other users. If Windows or Unix or Linux or Mac or BeOS could all do all of this, Linux would have a real shot - because it's free. Linux is highly reliable, but unfortunately that's about all it offers the typical desktop user.

To be clear: I'm not defending Microsoft. The courts have said the company achieved its dominance through illegal means. Microsoft has developed a well-deserved bully-boy reputation throughout the business world. As for its programs, Windows and Word sometimes drive me nuts. Why, then, am I interested in seeing Linux concede defeat on the desktop? Because I like Linux. I like the idea of Linux. I'd like open source computing to succeed.

Linux is at a turning point. Ten years ago, Torvalds first posted his OS on the Internet, inviting review, comments, and suggestions for improvement. Based on the principles of Unix, but written from scratch, its source code was open and available to inspect, use, and modify. A decade later, Linux is lauded as a technical success. But as a business, it's a flop.

No one expected to make money on free software. The idea was to make money offering Linux support and services. So far, not so good. No Linux company has figured out how to build a profitable business around open source software. In fact, several have already gone out of business and others are on the ropes. "Open source is a technology innovation strategy looking for a business model," says Eric Schmidt, now Google's CEO and chair.

What if the GNOME and KDE projects could take back all the programmer hours that went into consumer desktop applications and instead focus that brain power on developing kick-ass development platforms? What if all the effort that's gone into writing desktop drivers that peripheral outfits don't care to support were redirected toward drivers for corporate environments? What if all the mental energy, the rage on the Slashdot message boards, had been concentrated on building solid business models? What if the Linux community put an end to all the desktop nonsense right now and built on its strengths in global enterprise computing - just how big could Linux get?

What if all the mental energy, the rage on Slashdot message boards had been concentrated on building solid business models in enterprise computing? Just how big could Linux get?

In his book Innovation and Entrepreneurship, management guru Peter Drucker writes, "If an innovation does not aim at leadership from the very beginning, it is unlikely to be innovative enough, and therefore unlikely to be capable of establishing itself." Linux zealots might take that passage as their battle cry. But a more telling quote from the same book offers innovators this admonition: "Don't splinter, don't diversify, don't try to do too many things at once."

And no one should need Peter Drucker to tell them that playing catch-up against a dominant industry leader, in an existing product category, is a loser's game. Duplication is not innovation. If Linux had a chance on the desktop, established software vendors would be making apps that run on Linux. Instead, the open source community finds itself writing copycat versions of popular applications - the office suites, for instance.

Linux has been on the industry's radar screen since the mid-'90s, yet the vast majority of applications available for Windows and Mac don't exist for Linux. Microsoft apps - Word, Excel, PowerPoint - don't work on Linux because, of course, Bill Gates doesn't want them to. Nor does Internet Explorer, which controls 87 percent of the browser market. Linux-friendly Netscape Navigator, its market share in steady decline, holds a precarious 12 percent. That could be because Netscape was crushed by an illegal monopoly. Whatever. Those are the facts. Konquerer, an open source browser, doesn't even track market share figures.

Adobe Photoshop on Linux? Nope. QuickBooks? No. Eudora email? Uh-uh. Sega games? Nyet. And on and on. But Oracle, Sybase, SAP, Novell, IBM, and more see a market in enterprise Linux: They are busy porting their applications for server-based Linux.

The Linux desktop offers very little that could be considered plug-and-play. Linux drivers, the software that connects a computer with peripherals like printers and CD burners, are in short supply. Want to use a digital camera? Don't bother with Kodak if you're running Linux. Iomega is a bit friendlier, offering drivers for 14 of 51 products listed on its drivers Web page - but that compares to 42 Mac drivers and 45 for Windows. UMAX scanners? Not from UMAX, though some third-party drivers are out there, if you have the time and the will to find them. Creative's Soundblaster? A few are available, but you'll have to hunt down the company's software developer Web page.

Nontechnical users continue to have a hard time installing Linux. Every time a review appears in the press describing how damn hard it is to install Linux, and what few useful apps there are, it hurts Linux's reputation among top executives who might be considering the OS for enterprise.

Microsoft and Apple ads are everywhere, but no one is funding major marketing campaigns for desktop Linux. No one with any clout is carrying the torch for desktop Linux. Who is Linux's Bill Gates or Steve Jobs? Not Linus Torvalds. He supports desktop Linux, but does little proselytizing.

The movement's ideological focus is perhaps best expressed by Eric Raymond. A longtime Unix coder who embraced Linux early on, he wrote the seminal essay about open source development, "The Cathedral and the Bazaar," that spawned an influential book published in 1999. Raymond is the open source movement's self-styled Minister of Information. He speaks for a large population of Linux adherents who insist that the community bears not just a competitive imperative, but a moral obligation to bring Microsoft down. Says Raymond: "It's important that the money and power and market mass not be in the hands of outfits like Microsoft, who would destroy us."

That mission, he believes, requires open sourcers to stage a frontal assault on the desktop. "We've spent a lot of time making Linux techie-friendly - powerful, configurable," he says, "but not enough time making it easy for Aunt Tillie to use." Why should Linux try to lure Aunt Tillie? "Because," Raymond says, "dat's where da money is."

Matthew Butterick, a former member of Red Hat management who ran Web operations from the company's 35-member San Francisco office, disagrees. He believes the Linux movement is misplacing its talents with the desktop obsession. Linux might be cheaper than alternative operating systems, he says, but that doesn't take into account the switching costs or steep learning curve for most desktop users. "The anti-Microsoft rhetoric is an easy target that distracts a lot of the rank and file from noticing who the real enemy is - Unix vendors like Sun," says Butterick. "A lot of effort thus gets put into Microsoft-wannabe applications like the GUI, the office suite, the email client, et cetera. Some of the key features that would enable Linux to compete against corporate Unix are not getting done."

If the desktop is too appealing to ignore, Butterick thinks Linux application developers should focus on administration and software development tools for technical users. "Wouldn't a competent integrated development platform be more useful than an office suite?" he asks. "Wouldn't a graphical debugger be better than a whizzy file manager? Wouldn't a pro 3-D package be more targeted than an Outlook clone?"

Eric Schmidt agrees: Linux needs to attract widespread use of office and financial software to drive adoption, he says, but "Microsoft is extending Office in many ways, so it's hard to see this approach taking off in the near term." Yet, he adds, Linux-based desktops can be "software development platforms without peer." Google is a Linux booster. It has more than 10,000 crawling, indexing, searching Linux servers. Its developers use Linux on their desktops. But the desktop OS for everybody else? Windows.

Steve Solazzo, who heads Linux marketing for IBM, says that while the company makes Linux available on desktop machines, "the majority of current marketplace activity is on server-based deployments, so that really is where we're spending most of our time." Corporate customers feel the same way. Sara Garrison, SVP of technology and development at Visa International, says, "The real promise is in enterprise applications, in scalability. That ought to be the real focus."

There are would-be white knights out there. Chief among them, IBM, which has committed $1 billion to Linux marketing and development. Big Blue could do for Linux what it did for Java.

As if the lack of applications weren't enough of an obstacle, self-inflicted public relations problems - in the form of bad behavior by vocal Linux proponents - are casting a dark shadow over Linux. Eric Raymond, agree with him or not, is a thoughtful representative of the Linux movement. So is Linus Torvalds, and the top contributors to his kernel project, and many Linux business leaders as well. Unfortunately, the mass Linux culture is infected with a nastiness that creeps past contentiousness and toward the demeaning and dehumanizing whenever anyone criticizes Linux.

The problem has caused even Rob Malda, the founder of Slashdot, to sound the alarm. Malda, known by his nom de net, CmdrTaco, can get down and dirty himself. So when CmdrTaco's own troops provoke his disgust, you know there's a serious problem.

"Why Linux Won't Ever Be Mainstream" is the name of the controversial article he posted on Slashdot in July. It attracted more than 1,300 comments in a matter of hours. His two major points: Computer companies have shown little interest in helping Linux succeed in the consumer market; and, that's due in no small part to the bad attitude of the in-your-face Linux subculture.

Malda's rant recounts his attempt to find a USB device driver on the Web that would link his new HP scanner with Linux. In the process, he made a "shocking discovery": People were posting nasty messages aimed at HP. The messages, he said, were "just plain mean. HP employees are called bastards and assholes." One especially delightful post: "HP seems to be smeeling &##91;sic ] Gates' asshole rather than coming out of it. Beware, HP, Linux is going strong and unless you recognize that and properly support your hardware under Linux, your are going to Piss in your pants one day." This is not an isolated case. "I see it on Slashdot all the time," Malda wrote.

Most open source veterans take rabid flaming in stride. But that's not necessarily the case among the legions of open source volunteers showing up fresh from companies like IBM, HP, SGI, Motorola, and Compaq - not to mention their bosses, many of whom reserve judgment on the very idea of open source software. Which is why Alan Cox - effectively the COO for the Linux kernel development project and the man Torvalds refers to as his alter ego - wants the business community to know that the kernel team is a small and separate group with little if any real connection to the ideologues and the flamers. "There is actually very little overlap between the people doing software development and ... the Slashdot people, the people saying, 'You suck, you should give us a free driver,'" says Cox. "We take these vendors as responsible business people."

There couldn't be a more crucial time for Linux developers to cultivate their ties to business. The Linux industry is in trouble. Desktop software maker Eazel went bust. Desktop software maker Corel, a major Linux booster, has significantly cut back spending on Linux products in favor of the Microsoft and Mac platforms, and is now practically giving away software for Linux desktop. Red Hat stopped offering revenue and profit forecasts, and VA Linux, which exited the hardware business after determining that selling at negative gross margins was no way to run a company, is still bleeding cash. "The track record of the pure-play Linux vendors has been so poor," says George Weiss of the Gartner consulting group. "The investors have essentially walked away and said, 'We don't believe in you.'"

To survive, these companies need to make inroads into corporate IT, where Linux is gaining ground. IDC says 20 percent of companies responding to a 2000 study were using Linux to support a database; 10 percent were using Linux for a major app, such as CRM - double the figures from 1999.

Why is it being taken seriously in enterprise computing? Linux shares essential DNA with Unix. Their modular designs make the functions for both systems relatively easy to modify - changes need be made only to discrete parts of the OS. But in the past 30 years, Unix has split into different flavors, most of them closed and proprietary. Sun, IBM, and HP all have their own version of Unix, each at least partially incompatible with the others. Linux, meanwhile, is close enough to Unix that any Unix programmer should be able to make the conversion without much trouble. The resemblance makes it relatively painless for software makers to port their applications from Unix to Linux, if they see enough demand to make it worth their while. And because Linux source code is naked for the world to see, users have far more flexibility in adapting it to their own needs.

Cost also works in Linux's favor. Once the quality demands are met, companies worry about how much their computer systems cost. Linux can be downloaded free. It can be purchased from distributors for cheap. Sun, IBM, and HP have made their versions of Unix proprietary, with applications that run on only one system. Linux is effectively a commodity and can be made to work on any hardware system. This loosens the stranglehold that proprietary vendors have on their customers. With a choice of hardware, enterprise customers gain more control, and save money.

In other words, in the enterprise, Linux has a real shot.

That's not to say any of this will be easy. Serious technical issues must be resolved, the biggest of which is scaling. Big iron vendors like Sun sell systems that scale to 64 processors per machine. Simply put, more processors means more work at higher speeds. But writing OS software to accommodate scaling is tough, and Linux has struggled to scale from four processors to eight. The quicker Linux can move up the ladder, the better its chances.

There are social challenges, too. For starters, convincing top management of Linux's worth. Too many CEOs and CIOs still consider Linux a hacker toy. It's common knowledge in IT circles that open source software like Linux and Apache is being run on the sly at many well-known companies. The IT managers keep it hidden, knowing that their bosses aren't ready to offer their blessing. Swaying the top tier is crucial for Linux to succeed.

There are some would-be white knights working on these problems, the most promising being IBM. The company aims to make good on the service and support opportunities Linux presents, and, by sheer size and industry clout alone, it's in the best position to do so. IBM's huge user base justifies its attempts to sell high-margin hardware, software, and services on top of Linux. And given that the company has already devoted $1 billion to Linux development and marketing this year - including the ubiquitous Peace, Love, & Linux ad campaign directed at the enterprise - it's easy to see how Big Blue could emerge as the unlikely hero of Linux's commercial success. Consider Sun's Java: It didn't take off in enterprise computing until IBM embraced it and legitimized it as a cross-platform tool.

Gartner's Weiss understands Linux's appeal to IBM. "I'd rather have a Unix operating system that has worldwide resources with Torvalds doing the work for me," he says. "Then I don't have this big R&D budget."

IBM's proprietary Unix OS, called AIX, is well behind that of industry leader Sun Microsystems. But if IBM can use Linux to beat back Sun, it has a shot at regaining its leadership position. The company's participation is decidedly nonideological. "We are going to remain a mixed proprietary/open source software company," says Dan Frye, head of IBM's 200-member Linux development team. "We will continue to make billions off proprietary software.... We are not going to be a pure open source company - ever." And IBM executives have made clear that their interest in Linux is server-oriented. Although IBM owns Lotus, there are no plans to develop Linux versions of Notes, SmartSuite, or any of its other client applications.

It's not clear whether IBM's gamble will pay off. The open source method has proved itself in the Linux kernel project, which for all practical purposes is made up of 100 or so core volunteers who have learned to harmonize their work with the demands of Cox and, of course, Torvalds - long known as the project's "benevolent dictator." But it's far from certain whether the entry of big corporations, each of which faces plenty of its own intracompany battles, can effectively move Linux into the mainstream.

Already, turmoil is being created within the companies as decisions are made about what to open-source and what to keep closed. The Power PC Linux development "tree," for example, recently lost a leader - Cort Dugan, who, with the help of business partner Victor Yodaiken, had been running the project for six years. "We found ourselves intervening in intracompany disputes," says Yodaiken, who with Dugan runs FSM Labs, which develops a real-time OS called RT Linux, in Socorro, New Mexico. "Motorola's got the 860 which is not compatible with the 8260 which is not compatible with the 603 and the 604. And dealing with IBM on this is like dealing with five or six different companies. You have to keep track of 50 different ports." The pair turned over duties to a former Linuxcare developer from Australia who now works for IBM. "I hope IBM will take the leadership role," says Yodaiken. "It's kind of in a chaotic state right now."

Frye's IBM team may be gung-ho about open source, and the top brass all the way up to Lou Gerstner may be all for Peace and Love, but that doesn't mean everybody in between will necessarily go along. IBM programmer Dan Streetman recently wrote to colleagues on the Linux kernel thread about some anti-open source sentiment he encountered at Big Blue. Streetman was trying to solve a simple problem last summer - getting an IBM PS/2 keyboard to work with Linux. He fought with members of his department to open-source the actual driver, but the driver's author refused, and so Streetman was told to submit what he considered an "ugly" patch, a hook into Linux that kept the keyboard code proprietary. And this kind of thing is happening at all the companies now contributing to Linux - IBM, Motorola, Intel, HP, et al. None have figured out, in any methodical way, how to decide which code they can open up and which should remain proprietary - even when it comes to drivers for simple devices.

Not that any of this is unusual or unexpected. Most internal software projects are rife with politics and infighting - and open source schemes are, by their very nature, even more so because they're laid bare for the world to see. Torvalds, who has dealt with many headstrong individuals and insurmountable problems over the past 10 years, believes all this will work itself out. "I've gotten used to people having their own agendas," he says.

Clearly, a lot of work is needed to make Linux enterprise-ready. The competition in enterprise computing is fierce. Sun will be tough to squelch. Microsoft is gaining ground. But the field is still open. Nobody's got a hammerlock. Microsoft is just another player in this market - and there's no reason Linux can't break through.

But desktop computing? It's time to face reality: Linux is a loser.