The Linux desktop experience is killing Linux on the desktop, Part II

My two complaints are that the keyboard isn't the sort that I like, it's of the chiclet variety, and they only offer shortscreens (16:9 rather than 4:3). On the first complaint, this is really just my preference on keyboards. I'm a dvorak typist and so I'm a bit picky about keyboards; from my point of view, every laptop keyboard is bad, but I'm holding out hope for Thinkpads. On the second complaint it seems, unfortunately, that no manufacturer makes 4:3 laptops in 12"-15" sizes.

I have one of the ultra-thin models "Lemu1" and the body is plastic while the inner chassis is metal. It reminds me of the build on an iBook G4. It does feel a bit flimsy, but this model is about the size of a MacBook Air and so I figure some of the flimsiness is partly just physics.

I use StumpWM: http://www.nongnu.org/stumpwm/

I personally think the whole "Linux Desktop" issue is a bit of a red herring. GNU/Linux (and, I assume, other libre Unixes) is the only widely-available OS I know of where it is even possible to use a graphical environment without the "desktop" metaphor, which I see as a huge step forward.

To me, the "desktop" is all about entrenching the various kinds of boring, repetitive work that computers are supposed to relieve. This insidious metaphor is implicit in everything about how desktop environments are used and developed: the way users are encouraged to scatter icons across their screen for fast access to files, instead of encouraging organization and fast access via searching; the presumed importance of "office" software that most users could replace with a typewriter and a calculator; the focus on making the environment "intuitive" from the first moment one uses it, at the expense of training the user to eventually learn to use their computer more effectively and efficiently.

I would much prefer to see the free software community put its efforts into building graphical environments that eschew the desktop metaphor, and instead focus on making it easy for the user to off-load work to the machine. Part of that task is making it easy for the user to discover what the components of the environment are, how they work, and how to recombine them in novel ways.

A good test for this, I think, is how well a graphical environment supports a task like data entry, because it's the kind of thing that just about every computer is going to have to do at some point, and it is often a mind-numbing task. It's often mind-numbing not because the task is boring, but because most "desktop" environments provide no way to offload the cognitive advances you can make while doing it. After a few records, you know you have to check that field C looks a certain way relative to fields A and B, or that it's faster to enter field E before fields D and B, etc.; but the software you're using provides no way to encode this information, so you have to remember this boring algorithm and execute it repeatedly yourself, instead of letting the computer handle the algorithm while you focus on edge cases.

The challenge is to build an environment that a user can offload the algorithmic aspects of their work onto, without requiring the user to be a full-on programmer, at least not at first.

Free software has already taken some significant steps in this direction. Tiling window managers, the OLPC's Sugar interface, and even to some extent programs like Emacs are examples. We're ahead of the game when it comes to building a new metaphor for working with a computer. We just need to recognize it.

As far as I can tell, the "problem" with Linux on the desktop is that there is no consensus that there is a problem.

The category of people who use Linux, but complain about the desktop experience, seems to be a vocal minority (of the already tiny minority of computer users that use Linux).

The overwhelming majority of us have moved on. My Linux desktop use is now limited to the once a year that I install a distro to force myself to use a different OS for a couple months, mostly to see what I'm missing. Turns out, I'm not missing anything, so after a couple months I go back to using a system I enjoy.

But here's the thing, the people who use Linux as their main desktop environment aren't crazy, or fooling themselves. They prefer it to the alternatives. They evangelize about it (actually something I wish everyone would stop doing, regardless of their OS allegiance). For them, it's the best way to use their computer for the things that they want to use their computer for.

So I'm not sure that there is a problem. With the exception of the small number of people who fantasize about a world-wide exodus from Windows to Linux on the desktop (who are certifiable, if they think there's a remote chance that will happen), I get the impression that most Linux users are quite happy with it (as well they should be).

I think maybe twelve years ago, if you were a full-time Linux desktop user, you were probably compromising on technical quality for an ideological preference; but I don't think that's been true for a long time.

There was this post by a professional driver developer who wrote Linux drivers for the hardware company he works for. The problem he said, (I'm now paraphrasing) is that the kernel APIs kept changing and he had to keep reworking his drivers. Even though the driver code may be in the source tree, in the end it is still his job to keep it working.

Linux may have POSIX outside, but its cause would be further advanced if they committed to stabler interfaces for driver developers.

I didn't do that at all. Linux is doing fine as a niche desktop, and my post doesn't say otherwise. The point is that it's going to stay niche in its present state. All the hundreds of millions of people currently not using Linux don't have a reason to use Linux, and will continue not having a reason.

Whether or not this matters to you depends on how far you want Linux to go as a desktop operating system.