Could a common installer be the answer for desktop Linux?

Standardization isn't the only kind of cooperation that could further the idea of a free desktop

There's no mistaking that buying desktop hardware for use with Linux is more difficult than buying Windows PCs. Even if you manage to avoid the "Microsoft tax" and buy a system without Windows preinstalled, you still have the additional hurdle of getting your Linux of choice onto it.

As I discussed last week, Dell is reluctant to back any particular desktop Linux distribution. Instead, Windows-free PCs ship from Dell preinstalled with a version of DOS. They might as well come with blank hard drives; your first step to getting up and running with a Linux desktop will be wiping the disk and installing the OS manually.

Michael Dell says he might be more likely to back Linux on the desktop if the various distributions would concentrate on achieving a more standardized common core, but that seems unlikely. Even if you could convince the Debian, Red Hat, or Suse camps to reverse technical decisions made years ago in favor of a new standard, current efforts such as the LSB (Linux Standard Base) really only cover a fraction of the technical support problems that a hardware vendor would be expected to deal with. They make life easier, but not by much.

Last week I suggested another option: Forget about it. As long as there remains a wide variety of choices in the desktop Linux market -- and choice is a good thing; we want choice -- major hardware manufacturers will be hesitant to get on board. So don't expect a Linux PC to come with the kind of one-stop service contract that you can expect from a Windows desktop. Learn to live with the fact that, to a certain extent, desktop Linux is going to be a self-service job.

But suppose it didn't have to be quite that hard? Suppose the various vendors could all meet in the middle to lend you a hand?

Instead of standardizing each and every Linux distribution on a common core, developers from various distributions could come together to standardize a new, single distribution -- but not some kind of uber-Linux to end all Linuxes. On the contrary, the sole purpose of this lean, stripped-down distro would be to get a fresh machine up and running with the Linux of your choice. It could come preinstalled on factory hard drives instead of DOS or, alternatively, it would be small enough to fit on a Live CD or even a USB keychain drive.

Most every Linux distribution has some means whereby you can install the OS over the network. So, rather than using CDs, let's make this the standard method of installation. When you boot this distribution I'm imagining, it would configure your network connection and give you a menu of Linux distributions to choose from, with screenshots and descriptions of each, and then automatically pull down your pick.

For enterprise users, this same distribution could include code to locate a Linux installation server on the local network (using the open Bonjour protocol, perhaps). Plug in a brand new machine, connect the network, switch on the power, and installation of your preferred desktop Linux distribution would begin automatically.

Want to get even more ambitious? Suppose the installer contained an algorithm that would generate a standardized hash code representing all the hardware installed in your machine -- sort of like how Windows XP's copy-protection scheme works. This code could then be sent to Linux vendors for verification, so that a user could know that a given distro would actually work on specific hardware before downloading a gigabyte of software. Who knows? Hardware vendors could even publish these hash codes in their advertisements.

Would what I'm describing be easy to create? Probably not. In fact, it would probably be difficult. But it might be a worthwhile exercise to try, no matter how much energy it takes.

I often hear that competition in the Linux market isn't a zero-sum game; that in the long run a win for one corner of the Linux market is a win for everybody. Michael Dell seems to be saying that this isn't so. He seems to be saying that until a single desktop Linux distribution emerges as the clear winner, all the distributions have to lose. Maybe somebody ought to work on proving him wrong.

Copyright © 2006 IDG Communications, Inc.