Columns

Growing pains

• By Tony Baer
• April 1, 2001

In some ways, the idealism surrounding Linux resembles Java's idealism a few years ago, when the technology was regarded as a way of uniting developers who were otherwise divided by platform. The resemblance extended to the slimmed-down nature of both technologies with the implication that, at the end of the day, who needed all those add-ons like pointers, native drivers or embedded browsers, anyway?

Yet there are profound differences in the way each technology is managed, namely Linux's open source vs. Java's community source process. Specifically, it's the difference between that of super developer vs. platform vendor playing the role of traffic cop.

The question is whether or not these differences can help Linux avoid the rough adolescence that has plagued Java. As Java assumed enterprise-class capabilities, it became ground zero for the Sun–Microsoft war. Microsoft's recent Java licensing settlement sealed Java's fate—not as the Switzerland of application development, but as one of the antagonists.

With the Java issue "settled," will Microsoft now train its sights on Linux? Yes and no. Microsoft's Steve Ballmer has pronounced Linux to be the company's greatest competitive threat. But, unlike Java, he didn't have a single place to point fingers. It was left to Microsoft business development colleague Doug Miller to trash the open source OS by claiming, "There isn't much value in free."

The fact that no single vendor controls Linux has so far deflected direct attacks. And the idea that Linux can tap an entire virtual world of developers has proven effective when it comes to things like bug fixes and clever innovations. Even Microsoft has started to open its source-code kimono.

However, if Linux graduates from Web server to enterprise transaction server, some distributions will inevitably grow more equal than others because solution providers cannot afford to support infinite versions and high-budget development projects generally demand vendor accountability—and viability. To some extent, consolidation is already a reality, and most platform and software vendors are standardizing around Red Hat, SuSE, Caldera and TurboLinux distributions.

Consolidation is likely to occur sooner rather than later, thanks to upcoming Linux 2.4 kernel enhancements such as 8+ processor SMP support, 64Gb memory address space and new storage management capabilities.

Will success spoil Linux?
Someone will supply the definitive operating system, but who will it be? When IBM's Internet guru Irving Wladawksy-Berger remarked that the organization was making more money on middleware—like WebSphere or MQSeries—than on operating systems, somebody popped the question about whether IBM would eventually exit the OS business. Obviously laughable in the short term, IBM's point people did not slam the door on the notion. We're tempted to ask whether the tea leaves might be similar for Sun Microsystems.

The evolution to the data center dictates changes in the way Linux will evolve. Yes, development will remain open source, but when it comes to enterprise mission-critical systems, operating system updates are not likely to come from somebody, somewhere, but from established platform providers with documented support and managed update schedules.

That issue begs the question "What remains open source and what is proprietary?" Today, the Linux kernel is quite compact, presided over by its creator, Linus Torvalds. However, numerous critical building blocks, from the user interface to code documentation editors, to Symmetric MultiProcessing (SMP) support and clustering, are covered only to varying degrees in the kernel. If the feature is an extension, will open-source etiquette continue to prevail? For now the answer is yes, as shown by the recent release of several Red Hat clustering extensions by SuSE.

How long will it be before this decision is formalized beyond the careful deliberations of one celebrity or the collaborative kindness of strangers? As enterprise customers demand enterprise-class functionality, will the existing process continue to be adequate, or are standards committees—and standards politics—inevitable?

No one challenged the idealism that fueled Linux's growth as long as the technology was locked in a Web server cabinet somewhere in a back room. But as the platform begins to assume some of the features expected of an enterprise operating system, will open source be spoiled by Linux's success?