Our open source expert foresees the future of Linux: By 2012 the OS will have matured into three basic usage models. Web-based apps rule, virtualization is a breeze, and command-line hacking for basic system configuration is a thing of the past.
Is it optimistic to expect that by 2012, command-line hacking for basic Linux system configuration will be a thing of the past? One can hope, especially for things like display configuration, which should be auto-detected and configured touchlessly. This is crucial if Linux is to make headway with regular users, although putting Linux on devices like netbooks, where the hardware is a predictable and controllable factor, should help.
If there's any system configuration issue that divides Linux devotees, it's package management -- how to handle the wealth of software installed in a given Linux distribution. It's probably unrealistic to expect the plethora of distributions out there to consolidate on a single, one-size-fits-all package-management system, especially since each distribution tends to be married to its particular package management system. That said, the use of something like PackageKit as a packaging-neutral front end for a distribution might make transitions easier.
Also, the Conary package manager project offers some possibilities that deserve broader adoption, such as the ability to download and apply only changes to a particular package. That saves on bandwidth, which in turn ought to be a bonus for Linux users in developing countries where bandwidth is at an extreme premium.
Virtualization in the Linux kernel -- either in the form of KVM or Xen -- will make it that much easier to run Linux side by side with any other operating system, either as a way to migrate non-destructively from an existing Windows installation or as a way to expand Linux's own native functionality (for instance, by running multiple kernels each tailored for different needs).
Another possibility is to allow Windows apps to run side-by-side with Linux apps, using a virtual machine as the container for the former, and allowing cut-and-paste functionality between the two systems. To the uninitiated, it'll look like seamless Windows on Linux -- which is what we have with Wine right now, but using a VM as a wrapper would provide that much more flexibility. Using VMs would also make it possible to allow a project like ReactOS to work as a Windows container under this scheme.
Linux On Servers
It's almost foolish to expect Linux's dominion on the server side will wane -- servers are where Linux has fared best, and all the signs point to that only becoming all the more the case. The real question is, in what form?
A major part of the answer lies, again, in virtualization: Linux's mutability allows for its use not only as a server platform but as hypervisor and container for other operating systems. That said, there's more than one way to do such things -- KVM and Xen are two major contenders, but both function in markedly different ways and are probably best suited to different types of work. Xen's best for running as close to bare metal as possible, but KVM lets a particular Linux instance function as a container for other OSes. To that end, over the next four years, the question won't be "Which one's the winner?" but "Who's using each for what?"
The difference between the Linux of four years ago and the Linux of today is striking enough -- not just in its diversity, but in the way it's consolidating its strengths as a server platform, an OS for portable devices and emerging hardware markets -- and as a way to make the most out of whatever else we see in the next four years, too.