February 2003
If Linus cared more about what happens outside the kernel, it might be a less useful operating system.
On last October's Linux Lunacy Geek Cruise, over a hundred Linux faithful got to hang with Linus Himself for a week. Although conversation seemed to run more often to children than to technology (Linus and I were both there with our families -- between us we have four kids under six), Linus talked enough techno trash for me to gather that the man's mantra consists of three short words:
I don't care.
There's a lot of stuff Linus doesn't care about. Anything in "user space", for example. Other operating systems. New non-commodity microprocessors. Fights over development methods. The whole "free vs. open" thing, and the Richard vs. Linus story other people keep trying to tell. The list goes on forever, since there's a vast world of technical and political stuff going on outside the one thing he does care about: the Linux kernel.
His opened his talk on the boat with this disclaimer:
I only do kernel stuff. I did user level stuff about 10 years ago--only because without it the kernel isn't usable. I don't know what happens outside the kernel, and I don't much care. What happens inside the kernel I care about.
By not caring he doesn't mean he has no opinions. Like the rest of us, he has plenty of those. Unlike the rest of us, however, he's a Major Figure whose opinions are given a great deal of weight -- even when he goes out of his way to remove gravity by disclaiming any interest in a subject.
That's why Linus made news with a number of zero-gravity opinions he offered in his talk on the boat -- such as why he doesn't like Intel's Itanium or Apple's OS X.
It's only natural to assume that strong feelings accompany strong opinions; but this seems to be less true for Linus than it is for most people, because he so often goes out of his way to explain how little he cares about stuff that might be interesting, but also distracting.
Politics is a perfect example. In that same talk on the boat, he said "To me all the politics is just amusement value. I don't care."
And because he doesn't care, Linux is a huge success. In fact, it's getting more huge every day. Another geek cruiser was Roland Smith, Directory of Global IT Operations at LSI Logic. He explained to me how LSI Logic was gradually converting to Linux, pretty much across the board, from fancy engineering workstations to desktops. It wasn't just that Linux was cheap and useful, but that it uncomplicated a great may things.
This was consistent with a shift in the direction of news about Linux that I began to sense around the time of the cruise. The OS was suddenly being taken seriously, and not just as a "threat" to Microsoft. It was becoming established as a mainstream operating system -- perhaps the mainstream operating system. And the reasons were purely practical. Linux is cheap and easy to deploy. It's about as uncomplicated and useful as an operating system can be. These virtues are old hat for the Linux community, but they're new hat for many of the world's suits, who aren't used to an operating system that doesn't obsolete itself as a matter of policy.
In one of the panel discussions on the boat, the subject of obsolescence came up. Linus pointed out that commercial software is based to some degree on a model that values obsolescence. You make money by obsoleting old versions with new ones.
Case in point. When Windows 98 came out, Bill Gates was asked about threats from MacOS. Gates dismissed the question and said the real enemy of Windows 98 was Windows 95. His goal, plainly, was to extract fresh revenues from his entire customer base. Apple clearly has similar plans with major new releases of OS X. For two decades customers have taken this imperative for granted. They had no choice. It was pro forma.
Linux lets customers choose an operating system that doesn't care to obsolete itself. That choice became much more interesting to a lot of suits last summer when Microsoft raised its licensing rates for Windows -- something the market certainly did not demand. In a down economy this put customers in much better position to entertain the Linux alternative.
At the kernel level, Linux doesn't have a commercial agenda. Its purpose is to be useful, period. If you can find a way to make money with it, fine. Linus and his kernel don't care.
Sure, some things do get obsoleted along the way. In his talk Linus explained how the 2.6 kernel will have a whole new device block layer. But even those changes are not being made for the sake of obsoleting anything. They're being made so the kernel will be more useful, in more ways, for as long as possible.
The inherent practicality of the Linux kernel extends upward through the countless choices it supports. By contrast it's hard to imagine Microsoft or Apple wanting to support multiple desktops or UIs by developers other than themselves. But that's exactly what Linux does. It supports those desktops and UIs by not caring about them -- any more than the Devonian bedrock under Manhattan cares about the subways that tunnel through it or the buildings that stand on it.
Dave Sifry explains how it all works:
By focusing on a strong separation between kernel-space code and user-space code, the kernel is more stable, and strong user-space projects increase momentum. For example, by reducing the amount of code in the kernel, projects like Samba have been able to innovate in a decentralized manner, and ro create more stable, feature filled code. No one has to post a patch to linux-kernel in order to get changes to Samba made. All of the major subsystems of Linux share this attribute - XFree86, GNOME and KDE, browsers, heck, even glibc (although the glibc example isn't as strong as the others).
We also don't have to worry about Linus creating hidden APIs to make Open Office better than Abiword, or Mozilla better than Opera, or KDE better than GNOME, no matter which of those Linux personally prefers.
Not-caring is the ultimate level playing field. And it tends to best support other level paying fields built on top of it. For example, while Debian is perhaps the least commercial of all the major Linux distributions, it has provided extremely practical commercial "solution" building material for the likes of Lindows and Xandros. If Debian were busy caring about those implementations, it might be a lot less practical.
"Transparency" is another old hat Linux virtue that's becoming a hot buzzword in the world of suits. How long before customers demand the same absence of opacity in their operating systems as they do in their accounting sytstems?
Hey, why care? It's going to happen anyway.
Doc Searls is Senior Editor of Linux Journal.