Linux: the devil is in the details

As initial public offerings (IPO), mergers and product releases for Linux make headlines on a weekly basis, an equally important question is circulating through the high-tech community: How can companies differentiate themselves in a world where everything is open?

So far, Linux vendors are doing what the medical community adopted long ago to provide better care to patients - specialization. For example, on the desktop there is HelixCode, Red Hat and Corel. Others focus on turnkey hardware/software, such as VA Linux, and clustering, such as TurboLinux. In the long run, users can expect to see the number of vendors and vendor offerings slim down as companies merge and die off, observers say.

There are at least 157 Linux releases available, according to Dan Kusnetzky, an analyst at IDC, a Framingham, Mass., market research firm. Kusnetzky says of the 157, only five to seven come from top vendors, including Red Hat, SuSe (of Europe), TurboLinux, Caldera and Corel. For the most part, Linux vendors will make their way by providing added value to users beyond simple Linux software. That value will be in the form of services, support or software specialization.

"A lot of the support we look for is in the drivers. In the case of [Hewlett-Packard], we had specific hardware that we had to match our operating system to," says Chris Hawk, president of Solid State Design in Denver, which does application development for e-commerce sites. "The primary difference between all [the Linux] vendors right now is in the installation procedure and what types of software are being installed, because oftentimes you need to look beyond the operating system and into the prebuilt software that handles specific enterprise needs." He says TurboLinux, which focuses on clustering, is a good example.

There are others as well. Red Hat's Power Tools allows for local and remote administration services for servers and networks, with features such as host information, routing information and information on what services are loaded when the operating system is booted up.

Users should also test, test, test, Kusnetzky says. "Even if you bought five of [the releases], it's still less expensive than going with any of the major proprietary offerings," he says.

A company looking to cluster its systems might turn to TurboLinux's TurboCluster, which lets customers cluster servers running Sun Solaris, Windows NT or Linux. The software features dynamic load balancing, service monitoring, automatic IP failover and a set of management tools. But not all of it is open source. To maintain a competitive edge, TurboLinux kept some of the code closed for the first six months after the launch of the latest version of its software, then released it to the open source community.

To some observers, TurboLinux's approach is a good model of how to give to the open source community while maintaining a viable business. To others, it deviates from the open source model, which evolved in an era when Linux was developed by a wide network of people who freely donated their time - an era before shrink wrap and IPOs. "A lot of these people see themselves as a computer priesthood," Kusnetzky says, "but the real world enters in at some point, and the real world is not a pristine place."

Pristine or not, Cliff Miller, president and CEO of TurboLinux, says the company's approach lets it compete fairly and still give end users software that is less expensive and on a much faster development cycle than other offerings.

"Linux must remain open source and we are dedicated to that premise, but there are value-added features that can provide diversity," Miller says. "That implies choice, and Linux really is about choice, because the best [Linux products], or maybe more than one will continue to do well, thrive or survive and those products that are inferior will dry up and blow away."

But Red Hat does not agree. According to Michael Tiemann, Red Hat's chief technology officer, any company that offers Linux software can take a simple litmus test to figure out whether it is doing the right thing. "If you can read, modify and share all of the source code then that source code is open source," he says. "And if we call it open source it is true open source. But if someone requires you to use nonopen source software in order to run what they are giving you, then it is not true open source. Right or wrong, the market will decide."

Tiemann points to competitor Caldera, which he says has forced users to utilize a proprietary installer; SuSe, which has its own proprietary file system; and TurboLinux.

Tiemann says Red Hat has its own answer - keep everything open. Its Piranha clustering software follows that model. Tiemann says Piranha, like some of Red Hat's other offerings, suffers from lack of attention because Red Hat spends so much time preaching the open source message that its products don't get highlighted as often as the company's general mission to promote Linux.

Promoting Linux is easier for some other firms. At VALinux Systems, which provides turnkey Linux systems, being vendor-neutral appears to be part of the company's religion. The company loads Red Hat, SuSe and TurboLinux onto its systems, and plans to offer Debian (another desktop release of Linux) later this year.

Brian Biles, vice president of marketing at VALinux, says as long as the Linux operating system kernel remains open and developers continue to make improvements, it will improve. According to him, it would not be good if Linux development follows Unix, which ended in separate flavors and a fragmented community. "Everyone got burned," Biles says.

So what's a network manager to do? Observers say making an old-fashioned checklist is a good approach. "If I am a network manager, I know what I want my systems to be - an FTP server, database integration, etc.," says John Dunkle, an analyst with Workgroup Strategic Service in Portsmouth, N.H. With that in mind, Dunkle and others say, network managers need to look at what vendors have to offer in the way of tools to manage their infrastructure.

In the end, Kusnetzky says, network managers should remember that trying out a few offerings might be worth the hassle, considering the dramatic savings of a Linux implementation, which can be thousands of dollars less for each client on the network compared to leading operating system implementations.