Why Linux Will Win
Sunday, June 27, 1999 by Dave Winer.
On Friday PC WEEK released the results of its benchmark tests between Microsoft Windows NT Server and Linux.
PC WEEK presents an inaccurate picture. They make it seem as if Microsoft is leading the web server market. Microsoft actually has problems to fix, and could be trailing in the areas that matter most to system managers and developers.
Here's a list of the issues I have with the PC WEEK benchmark.
1. Price-performance.They chose to run the two systems on identical hardware, but Microsoft's server software is far more expensive than the Linux software. Hardware prices have dropped so much that the cost of software is now significant. Would they compare spreadsheets without considering price? Further, what does NT Server actually cost? I looked for but couldn't find a pricing page on Microsoft.Com. This information is required for a meaningful comparison. It should have been included in the PC WEEK analysis.
2. Who cares about static pages? The only kind of site that runs in comparable volume to those in the test are super-high-volume sites like Yahoo or AltaVista or Microsoft.Com. But these sites don't run static pages now, if they ever did. Today static page performance is irrelevant. You can get great performance for static pages from hardware that is essentially free. The obsolete 486s and 68Ks of the past make excellent static page servers.
3. Dynamic performance is what matters. PC WEEK didn't compare the environment most system managers care about. How does it perform serving script-generated, database-stored pages? Ask your favorite webmaster if you don't believe me. This is where we're looking for performance.
4. Options. Linux is starting to swamp NT Server with different pricepoints and options. I have a Cobalt Qube, it costs $1000, it runs Linux, it's trivial to set up. There's versatility. Can you get it from Microsoft? Put another way, if I have $15,000 to spend on new server hardware, would I do better with Linux or NT? That's a fair question for a benchmark to address.
If PC WEEK wanted to help its readers they'd design dream systems for each operating system. With a budget of $5,000, and a full-featured site to run, figure out what configuration of each system would make such a site run beautifully on today's hardware. That would be far more positive than declaring one system a winner and the other a loser.
The editors at PC WEEK will hopefully remember IBM's OS/2 Presentation Manager. It was IBM's windowing system, co-developed with Microsoft, and introduced in the late-80s, while IBM was at the pinnacle of power. Even so, Windows won, it became the defacto desktop system, not Presentation Manager.
Linux will survive for the same reasons that Windows survived. It's more practical, the platform vendor is less powerful, and therefore the opportunity is greater for developers, and users will have more choices. Further, Linux has the advantage of being much less expensive than its competition and is free of licensing restrictions.