Home
About
Perspectives
Writings
Contact
Resources
WritingsPerspectivesCollateralMediaResearchOverview

Originally appeared in MSI Magazine
June 1, 2000

The Kernel’s Secret Recipe

Among other things, Colonel Sanders was known for developing a standard, secret recipe of 11 herbs and spices and a cooking method that could be replicated to the point where you could walk into any KFC shop and always expect your chicken to taste the same.

In many ways, the IT world has tried duplicating the late Colonel’s secret formula. Your company’s Windows is the same as somebody else’s. The main difference? Instead of 11 secret spices, you get hundreds of equally mysterious dynamic link libraries and other features. Try fixing a bad listing in the Windows Registry. In the end, you’ll probably decide that reinstalling Windows—with all its attendant disruptions—is the less risky course because of the perils of prying open the black box. However, compared to the days when every computing shop ran its own unique systems, that proved a great leap forward.

Does the sudden popularity of Linux question the notion of enlightened secrecy? Linux has been built using the “open source” model of software development that encourages users to open up the black box and toy with it, as long as the changes are shared with the rest of the community. On the surface, that sounds like chaos. Lot of Linux versions, changing all the time, lots of fragmentation all around you. In other words, the UNIX mess all over again.

On the contrary, says Linus Torvalds, the inventor of Linux, and the person entrusted with the OS’s kernel. He says that fragmentation can be good as long as it comes from users adding valuable features, rather than vendors adding extensions for locking in their customers. Today, Linux consists of a kernel whose release schedule coordinated by Torvalds, and countless extras introduced by users and vendors alike. Those extras are the things that differentiate the Linux distributions coming from commercial sources like Red Hat, Caldera, Corel, or SuSE.

Until recently, the mainstream IT community dismissed Linux as the little operating system that could. Used mostly by backroom developers for performing extremely specialized tasks, such as running small websites, nobody in their sane mind would dare running Linux for more mission-critical applications. Among the drawbacks, the current 2.2.x version lacks scalability, essential features for managing application recoveries after crashes, and consistent support of libraries for popular programming languages such as C.

And, as we pointed out a year ago (see Manufacturing Systems, March 1999), relying on an operating system that nobody owns—or takes responsibility for—can be a career-limiting move if your server crashes at 3 am.

But, as we also noted last year, household name IT vendors were on the verge of offering support packages to close the uncertainty gap. Today, that’s reality, with IBM, Dell, and Compaq bundling versions of Linux from specific vendors such as Red Hat and Caldera on their machines, along with support packages.

Why are all these big names embracing Linux? It’s the reliability, stupid. Linux’s modular structure—which lets you pick and choose only the features your application needs—has become a powerful selling point because it reduces the impact of bugs caused by bells and whistles that you didn’t necessarily need. In other words, the exact opposite of Windows.

Linux is great for deploying in places which the IT folks visit maybe once or twice a year. No wonder that Lotus is pushing the Linux version of Domino to replicated applications, such as inventory tracking systems deployed by retail chains at their individual stores.

Therefore, when considering Linux, ignore the purist hype. The open source debate is relevant only to vendors—who must decide which versions to support—and hackers—who like staying after hours to download and play with the latest code.

While the latest Linux releases circulate through the open source community, enterprise Linux users will remain bound to more conventional vendor bundle and support contracts with controlled release schedules. Upgrade decisions will be akin to deciding whether to take SAP’s latest R/3 version or waiting for the next major dot release. To mainstream Linux users, the kernel’s secret recipe may remain a black box, just like Windows. But as long as they get that 3am support, who cares whether Linux is open source?


all original content copyright © 2001 - 2006 onstrategies
site designed and copyright © 2001 - 2006 oz barron