All sorts of theories have been proposed about why Linux has not taken a significant chunk of the market for desktop OSes. While there may be merit in some of the psychosocial economics presented, it ignores big, more concrete reasons why Linux hasn’t seriously eroded Windows market share, and why even Mac OS X is outstripping it on the desktop front.

#The pre-installation argument

Everyone knows that people are generally lazy. If your machine comes with XP, you’re probably going to just use XP. Even when they only have the base OS installed on an otherwise empty hard drive, very few people are actually motivated enough to wipe out their hard drive and install a different OS. (And the longer someone gives in to inertia, more and more crap will accumulate on the hard drive. And the more cruft that accumulates on someone’s hard drive, the less likely they are going to want to reformat and reinstall.) This is what makes the rebellion against using Vista so remarkable. It’s the first time people are demanding that they have the right to delete the base OS and install something else.

But notice something about the Vista debacle: people are demanding that the computer manufacturers themselves should be responsible for pre-installing XP. They don’t want to deal with having to do the reformat and reinstall themselves.

Now I know that Dell offers machines pre-installed with Ubuntu, but it’s not the default. You have to ask for it. Even this small obstacle is probably enough to deter people from adopting Linux. The fact that you have to ask for XP to be pre-installed instead of Vista is probably the reason that Vista’s numbers are even as high as they are.

Pre-installation is the reason why IE ended up destroying Netscape. Despite the fact that IE 1 was worthless and IE 2 was still rather primitive compared to their Netscape counterparts, the fact that you didn’t have to download and install anything essentially assured IE’s dominance.

#Hardware first, software second

It’s true that most Linux distros have all the capabilities that most casual computer users need from a computer: web browsing, basic office productivity software (word processing, spreadsheet, graphics manipulation), and the ability to listen to mp3s and watch movies. The most popular Linux distros sport desktop interfaces that are decidedly Windows-like, and most computer-literate people have no real problem with navigating the Linux desktop.

But the average person doesn’t really choose an OS. They decide what kind of computer they want to buy.

In the same way, people don’t really choose automatic transmission or four-wheel drive. They decide what kind of car they want to buy.

And if almost every single computer manufacturer out there pre-installs Vista on their machines, then more likely than not, unless you’ve got very particular needs, you’re going to end up with a computer with Vista.

Notice that Mac users don’t choose to run Mac OS X. What they decide is that they want to buy a Mac. And not just any Mac. They have to choose between very different form factors: MacBook, MacBook Pro, MacBook Air, Mac Mini, iMac, Mac Pro, XServe. While in terms of architecture, these machines are interchangeable, their specific hardware is very different, in the same way that while two cars may have the same engine and chassis, the sports coupe is decidedly different from the station wagon.

The choice of hardware is what is going to cost you big bucks. And you just have to live with the OS it comes with.

So until a particular computer manufacturer has a particular stake in selling Linux machines, and until they start exerting marketing resources to that goal, desktop dominance just ain’t going to happen.

#Disconnected from the goal

I think that the reason why Mac users are so fanatical about their OS, but more importantly, about their machine, is that they know that Apple is interested in designing machines specifically for the home desktop. So Apple’s strategy has always been to design the hardware first, and the interface comes as a natural result. Since their goal is to sell hardware, they need to make decent software to make buying and owning a Mac a worthwhile experience. So they’ve research optimal UI design, and they’ve come up with the Human Interface Guidelines, and most popular applications on the Mac are extremely intuitive to use.

In contrast, the average manufacturer of x86 machines don’t really specifically target the home desktop. The same Dell that sits in your house is probably the exact same Dell that is sitting in the office, except maybe the one in the office doesn’t have speakers and has a bigger hard drive. And in reality, most manufacturers of x86 machines care way more about the business market than they do about the consumer market.

As far as Mac OS X goes, Apple isn’t so much interested in creating an optimal OS as they are in creating an optimal interaction and experience with the hardware they want to sell you.

In contrast, Microsoft just wants to sell software. In the 8-bit days of yore, they didn’t even sell code directly to the consumer. They contracted with computer manufacturers to design their OSes, or more accurately, their interfaces. Back in those days, the interface was generally a BASIC interpreter. And Microsoft wrote BASIC implementations for all the major players in those days: Apple, Atari, Commodore, Tandy, and IBM.

The first actual OS (or more accurately, kernel and command-line interface, I suppose) that Microsoft built was MS-DOS, initially licensed to IBM (who customized it and released it as PC-DOS), but then became available for other OEMs to license.

So all of these companies took Microsoft’s source code and customized it for their particular machines. Whether or not a machine sold wasn’t Microsoft’s responsibility. And platform compatibility was something no one really cared about.

When various computer companies managed to reverse engineer IBM’s BIOS, the age of clones had arrived, prices dropped, and everyone pretty much standardized on IBM-compatible architecture, which meant an x86 CPU with BIOS and MS-DOS running on top of it. When Compaq released the first 32-bit machine before IBM did, this set the stage for the first commercially successful version of Windows: 3.0.

The rest, as they say, is history. The power to design computer systems was wrested from computer manufacturers, instead becoming dictated by the ability of a computer to run Windows. Everything degraded to the lowest-common denominator (although there were in fact many computer manufacturers that dared stray from the beaten path, all of whom died lonely deaths, even including IBM’s personal computer division). And Microsoft really doesn’t care how particular computer manufacturers do. Even if Dell and HP went bankrupt, other companies would certainly take their place, and Microsoft still would continue to make tons of money.

This is, unfortunately, the landscape in which consumer-grade Linux finds itself: a market filled with indistinguishable x86 machines that are first and foremost Windows-capable.

In reality, Linux is generally successful in the realm of server-side machines, running on a wide variety of architectures. And the server market is very divergent from the desktop market. Heavily used servers often even eschew GUIs in order to maximize CPU cycles available to high-availability apps. UNIX and UNIX-like systems basically dominate this market. And ultimately, Linux is a UNIX-like system best suited to be a server, or to function in a cluster to run computation-intensive but often non-graphical applications. This is the market to which Linux-based companies like RedHat, Novell, and Oracle.

#Too much imitation, not enough innovation

With regards to the desktop, however, Linux is essentially playing catch up. To the unfamiliar eye, both KDE and GNOME superficially look decidedly Windows-like. And most casual users tend to treat these environments as Windows knock-offs.

While features standard to X-based GUIs for decades which Windows never implemented are preserved, such as virtual desktops and multiple viewports, most users who are familiar with Windows don’t really know what to do with them and tend to ignore them.

In fact, hard core Windows users even tend to be baffled by Mac OS X’s intuitive interface.

True, Windows is not the only interface that X on Linux can emulate. But one of the other major desktop environments, GNUstep, is in fact a NeXTstep-clone (NeXTstep being the direct ancestor of Mac OS X)

Until Linux desktop environments develop truly useful features that Windows lacks, or alternative interfaces that transcend the dominant desktop metaphor, most people will continue to think of Linux as a Windows knock-off, and seen in this light, Linux will never look favorable in comparison.

#The road ahead

While I’m not saying that Windows-like functionality should be ripped out of Linux distros, there needs to be a focus on alternative interfaces. Look at the direction that Apple is taking. The iPhone’s and iPod Touch’s interface is likely to be adapted to notebooks and desktop machines, bringing us closer to the fantasy interfaces found in “Minority Report” and “The Matrix”. And the Wii’s accelerometer is just waiting to be used in non-gaming applications.

Until Linux distinguishes itself as a unique desktop experience and not just a reimplementation of Windows functionality, it’s unlikely to make any in-roads on the desktop front.