John Halamka has a penchant for experiments with new technologies. In 2004, the now 44-year-old CIO of the Harvard Medical School and CareGroup, which runs the Beth Israel Deaconess Medical Centre, who is also a practicing emergency room physician, was one of the first people to have an RFID chip containing a link to his medical records implanted in his body (it's near his right triceps.) In April, he and Harvard geneticist George Church will become the first humans to have their DNA sequenced and their full genetic makeup posted on the Web.

But as a health-care administrator, he's not solely interested in testing the cutting-edge, Orwellian technologies that make headlines. The PCs inside the hospital have to work too. So when Halamka's laptop running Windows XP interrupted several presentations with inopportune antivirus and application updates, he decided his next big initiative would be to determine which desktop operating system — Windows XP, Apple's OS X or Linux — is the most secure, most reliable and easiest to use in a corporate environment.

For three months, Halamka ditched his Windows laptop. He replaced it first with a MacBook running OS X. Then he spent a month using a Lenovo ThinkPad X41 running a dual-boot configuration of Red Hat Enterprise Linux Workstation and Red Hat Fedora Core. Finally, he took up a Dell D420 sub notebook running Microsoft's Windows XP. After evaluating all three to determine which worked best for him, he plans to begin testing his preferred setup with users, most of whose desktops currently run Windows.

Halamka judged the three operating systems according to a variety of criteria including their performance, user interfaces and enterprise management capabilities, such as the ability to configure applications, easily organize file systems, and establish granular security control. We followed Halamka's progress, and now we have his conclusions. We've also ask three other experts to take a look at Halamka's findings and add their own insights.

Halamka admits to a bias against Microsoft: He thinks the complexity of the Office product suite hampers its performance and makes it more vulnerable to viruses and spyware. Halamka says that at a dinner with Microsoft CEO Steve Ballmer in 2004, he told Ballmer that he doesn't use 80 percent of the features in Office and suggested that Microsoft develop a simpler, more secure and reliable product (Ballmer's response, he says was to cite statistics that indicated lots of users like 95 percent of Office's features). Halamka is also sceptical of Microsoft's future in creating simpler, more reliable products, given Bill Gate's upcoming retirement and the company's appointment of CTO Ray Ozzie to succeed Gates as chief software architect. (See our story "Beyond Vista" for more on Microsoft's future goals.) "Ray is brilliant, but his two products, Lotus Notes and Groove Networks, are both huge systems that taxed networks. So what's the likelihood of Microsoft coming up with a simple, reliable product suite? Pretty low," Halamka says. Meanwhile, Halamka notes the rise of Google and Linux as credible challengers to Microsoft's dominance and Apple's adoption of Intel chips as a way to boost the performance of its products and enhance their potential to be used in a corporate environment.

People are risk averse. They don't like to try new things. It's inertia that holds Microsoft in place

John Halamka — CIO Harvard Medical School and CareGroup

Few companies are seeking alternatives to the Microsoft desktop, according to Roger Kay, president of IT research company Endpoint Technologies Associates. "People are risk averse. They don't like to try new things. It's inertia that holds Microsoft in place," he says.

Even organizations that are seriously investigating Linux and Apple tend to keep their evaluations on the QT, according to Rob Enderle, an IT analyst. Enderle says that public companies don't like to publicize their experiments because if their attempt to switch from one operating system to another fails, they want to be able to quickly and quietly sweep the project under the rug — lest they draw negative attention or jeopardize their relationships with existing vendors.

Halamka tested the operating systems himself before testing them with users because he wanted to know firsthand what problems users might encounter and get a sense of whether his IT department will be able to easily and cost effectively maintain the platform. He conducted the experiment before the release of Apple's Leopard and Microsoft's Vista operating systems for two reasons: He had the time in his schedule to learn the nuances of the different operating systems, and he prefers testing established, stable technologies rather than new releases.