Linux Takes Off -- But Where Is It Really Going?
After a slow gestation, the free operating system is quickly maturing. Its ultimate fate, however, is far from certain
When Brett Thomas, technology vice-president for online music retailer Goodnoise.com, set up a server at a local Internet service provider last May, his contact at the ISP asked for Thomas' pager number so he could be contacted when the machine crashed. The assumption was that Goodnoise would be running Microsoft's Windows NT operating system on the server, which stood to get some heavy loads as surfers downloaded music files in the popular MP3 format. But when Thomas told the contact that Goodnoise was using the Linux operating system, he replied: "If you're using Linux, I don't need a pager number."
Thomas gave him the number anyway. But he says Goodnoise hasn't yet suffered a software-related crash even though its traffic has increased dramatically since the deal was done. And Goodnoise uses Linux, a free version of the Unix operating system that runs a large share of the servers on the Web, for a lot more than delivering music files and Web pages. The upstart operating system also manages several internal Goodnoise networks, including the company's E-mail system and several desktops.
Thomas isn't alone. Information managers throughout the corporate world are discovering that Linux is a viable challenger that often outperforms Microsoft's NT operating system and other versions of Unix -- at a fraction of the cost. Last year, according to market researcher International Data Corp., there were more than 750,000 new installations of Linux, giving it 17% of the low- to medium-size server market. If Linux' growth rate continues, its share of that $2 billion segment of the server market could double by yearend, eclipsing Microsoft's NT as the segment leader.
--------------------------------------------------------------------------------
Free software is a deal for companies used to spending millions on operating systems
--------------------------------------------------------------------------------
Clearly, Linux has come a long way from its birth less than a decade ago as part of a Finnish student's doctoral dissertation. It has quickly gained converts thanks to its two "free" characteristics: It is free of charge. And, unlike the case with commercial software, programmers are free to make changes to the program's source code to customize it to their needs -- as long as they make those changes publicly available. That so-called open-source element is a deal for corporations accustomed to paying plenty for operating systems such as NT or one of dozens of proprietary Unix variants. A large multinational, for instance, may normally spend millions of dollars on operating-system software.
Better yet, because Linux is an open-source program, a systems department overseer can quickly tap into the source code and write a software patch to fix a problem. If the network instead runs on a proprietary operating system, the overseer will have to reach the customer service department of the software maker, then wait for its programmers to write a software patch (since the customer isn't allowed to see the underlying code of the program) -- then hope that the patch works. "Linux allows your company to run software, not a software company to run you," says open-source advocate and self-described hacker Eric S. Raymond.
BOLD CLAIMS. So far, all that those advantges have bought Linux is a legitimate chance to play in the game. Linux currently owns only half the market share that NT does, and much of that represents small servers doing "robot" tasks -- the same one over and over again. While NT has so-called scalability problems, moreover -- meaning that it has difficulty handling heavy loads on big, multiprocessor machines common in corporate setups -- other versions of Unix, such as Sun Microsystem's Solaris and IBM's AIX, have so far proven more scalable than Linux, and thus are more trusted to run enormous, sophisticated networks. Even so, Linux' cheerleaders aren't afraid to make bold claims. Raymond predicts that the software will be on more than 50% of the world's computers within five years. "And that's a conservative estimate," he says. "The battle between Linux and NT is already over except for the shouting."
Microsoft, whose NT system runs on more than 25 million computers today, understandably sees things differently. "[Open source] is more of a public relations phenomenon," says Jim Ewell, marketing director for the next generation of NT, called Windows 2000. "I don't think we or our customers need open source right now." Ed Muth, the group product manager for Windows 2000 echoes that opinion: "Linux is just another brand of Unix, and I don't think the world is crying out for another Unix operating system."
--------------------------------------------------------------------------------
Big shots from IBM to Sun to HP have thrown their weight and financing behind Linux
--------------------------------------------------------------------------------
Their arguments would be persuasive, except for one fact: More and more corporations are adopting Linux, at the expense of both NT and commercial Unix. Indeed, just about every owner of a proprietary Unix operating system, from IBM to Sun Microsystems to Hewlett-Packard, has thrown its weight and financing behind Linux by making investments in the leading Linux companies, including Red Hat Software, Caldera Systems, and VA Research. As the program gains critical mass, more and more applications programs are being written to run with it, making information managers more willing to give it a try.
The story of Linux starts at Bell Labs in the early 1970s, where a handful of engineers created Unix to aid in their research. Since AT&T saw no commercial potential in the software, it gave the program's underlying source code to anyone who wanted it. Unix quickly became popular among academics, and by the time the PC revolution erupted in the early 1980s, Unix was one of the most widely used operating systems for running networks of computers.
But there wasn't just one version of it. Corporations such as IBM and HP quickly developed their own flavors of Unix, all of which soon became noncomplementary, meaning you couldn't run the same application on different versions of Unix. By the early 1990s, Unix was still the predominant operating system for servers -- computers that run networks -- with more than 30 variations.
LINUX ISN'T UNIX. Enter Linus Torvalds, a graduate student from Finland, who in 1992 decided to write his own version of Unix for his doctoral dissertation at Helsinki University. His operating system kernel -- the most basic element of an operating system, without any device drivers or other programming that makes a computer run -- resembled the original Unix in many ways except in name: Linux, stands for Linux Isn't Unix.
Coincidentally, a group of programmers in the U.S. calling themselves the Free Software Foundation was putting together a Unix-style operating system called GNU (which stands for Gnu is Not Unix). That group had all of the working parts of an operating system except for the kernel. When Linux met GNU, it was a perfect match.
--------------------------------------------------------------------------------
The Linux license helped the program take an unguided, evolutionary development path
--------------------------------------------------------------------------------
Linux got more from GNU than its programming. It also inherited its novel terms of agreement, called the General Public License (see "Linux May Be Running on Some Spindly Legal Legs"). Written in relatively plain English, the GPL allows anyone to view and alter the source code of Linux as long as the alterations are made public. That has helped the program take an unguided, evolutionary development path, far different from one dictated by the focus groups and team meetings that led to the development of other operating systems such as Microsoft's Windows or Apple's Mac OS. The arrangement also caused GNU and Linux to develop very slowly in their first year or two, since it was very difficult to share programming changes when you had to schlepp a floppy disk around to show your buddies.
At that point, however, the perfect communications medium appeared -- the Internet. Word of the new operating system spread, and Linux began to win converts worldwide. By 1996, it was installed on several thousand computers and was being used for real tasks, such as serving pages on Web sites. The next year, it won trade magazine InfoWorld's award for "Best Operating System," and for the first time the business world noticed.
But it wasn't until 1998 that Linux became a household word. And many experts think that in 1999 it will be adopted by corporations the world over as their standard server operating system. The list of companies that have aligned themselves with Linux by making investments in Linux companies or by throwing programmers into Linux projects reads like the batting lineup for a technology industry all-star softball game. In addition to IBM, Sun, and HP, they include Compaq, Dell, Intel, Oracle, and SAP.
CODING DNA. Will Linux gain the type of popularity in corporate circles that it has won in academic and programming communities? It sure could, if you talk to corporations that have already experimented with it. Burlington Coat Factory, for instance, recently ordered more than a thousand computers from Dell Computer Corp. with Linux preloaded for its point-of-sale system. A large E-commerce Web site (whose officials won't comment publicly because the company is in a Securities & Exchange Commission-mandated quiet period in advance of its initial public offering) is run completely on Linux and reports that it suffered only a few minutes of downtime during last season's Christmas rush.
Or take, Incyte Pharmaceuticals, which does DNA sequencing on a massive scale. It has hooked dozens of Linux computers together to do complex calculations that sometimes take up to three weeks of nonstop computing. "You can imagine how upset we would be if the computer crashed in the middle of that three weeks," says Vice-President Steve Lincoln. "We're not religious about this. We chose Linux because of our positive professional experiences with it, and it has performed without a hitch." A year ago, Incyte had no Linux on its computers. Now, more than a third of them run it, and that number grows daily. Lincoln is looking into transferring other computer operations within the company over to Linux, including its E-mail server and possibly even its network of desktops.
--------------------------------------------------------------------------------
There are three primary problems facing Linux
--------------------------------------------------------------------------------
Such success stories are becoming common. But before you start writing off Windows and selling your Microsoft stock, it's important to remember three things:
Linux was designed to be a server operating system. While it's possible to use it as a desktop OS, there's little reason to do so. For one thing, it -- like Unix -- is so difficult to install on a PC at the moment that the average person (including me) can't do it.
Few applications have been written for Linux so there's very little you can do with it once you have installed it. A few word processing and spreadsheet programs have been written to run with it, but for the most part Linux' desktop uses are limited.
Linux lacks a standardized GUI, or graphical user interface. Without a GUI, you have to communicate with your computer via "command lines," much like with DOS, Microsoft's predecessor to Windows. Several GUIs are available for Linux today, but it's a nightmare to get them to communicate with the peripherals that go with your desktop, such as your printer or graphics card.
All three of these problems should eventually be solved by the armies of programmers who are constantly improving and solidifying Linux. But for now, it isn't a threat to Microsoft's Windows or Apple's Mac OS as a desktop operating system.
This doesn't mean Linux is going to turn into a sideshow, however. With the fantastic growth of the Internet, computer networks are quickly becoming more important. Even if Linux can't conquer the desktop, it may win the computing war by becoming the dominant OS for servers. Eventually, most desktops will run better if they run on the same OS that's running the network they're on, and in the three to five years it will take to determine a winner among server operating systems, Linux should improve to the point where it will be a true competitor to Windows.
NOT STANDING STILL. "Just a year ago, I couldn't use Linux for basic office productivity tasks," says Greg Hankins, the senior network engineer at MindSpring, a large ISP that uses Linux on everything from several internal servers to some of its desktops. "Today, I can read Microsoft Word documents on Linux. A lot of people are working on a lot of different applications that are going to make it a very usable desktop operating system in a very short period of time."
Of course, Microsoft isn't standing still either. It's working on features such as continuous voice recognition, multiple language interfaces, and even face recognition for the next version of Windows. One of Linux' biggest challenges will be to stay on the cutting edge without centralized leadership to give it direction. Eric Raymond sees that as an advantage, however. "Linux' dispersion of authority is a virtue," he says. "It means that there's a fluid market in improvements in Linux all the time."
Many believe that the greatest threat to Linux is the tendency of free software to fork into different versions. "Forking is one of those problems that Linux critics have been predicting since the beginning," says Bob Young, the CEO of Red Hat, the leading for-profit distributor of Linux. "It's a problem that hasn't happened, because of the nature of open-source development. But if it does, it could be a problem." Look no further than the original Unix for an example of forking. Without an authorized owner to decide what goes in and what stays out, Linux could quickly develop into multiple operating systems that don't work together, much as Unix did.
--------------------------------------------------------------------------------
"Since no one can own Linux, there's no reason to make an incompatible version"
--------------------------------------------------------------------------------
Still, there's a key difference between the original Unix and Linux: the latter's unique user license. Raymond thinks that because it's impossible for Linux to be turned into a proprietary product, there's no motivation for forking. "The reason closed software forks into incompatible versions is so that one company can differentiate its version from everyone else's," he says. "Since no one can own Linux, there's no reason to make an incompatible version. Anybody that does do that will see their product quickly die because there's no reason anyone else would want to use it." The fact that Linux has not forked in its first seven years of existence is a sign that it may not in the future. But the possibility can't be ruled out.
Whatever happens to Linux itself, it could lend a major push to the fledgling movement toward open-source, free software. In some ways, open source has been around for quite awhile. "Open standards are what allowed the Internet and E-mail to explode," says Arthur Tyde, CEO of LinuxCare, a company that provides service to Linux users. "Now let's see what will happen when an operating system is based on open standards."
There's no lack of other types of mass-market, commoditized software that could copy Linux' open-source model. Already, traditional software companies such as Netscape and Apple have adopted the open-source model for key software lines, including Apple's server operating system and Netscape's next-generation browser (see "Don't Close the Book on Netscape's Open-Source Browser"). If Linux pans out as predicted, it could set a precedent that other software companies might be compelled to follow.
By Sam Jaffe in New York |
