Most distributions today can be used for anything - a desktop system, a web server, a file server, a firewall, DNS, firewall, etc. I am of firm belief that Windows' greater downfall on the server is that it has been a glorified desktop for too long. The file extensions are still hidden by default, you're forced to run a GUI, and you can still run all your applications on the system. I predict that we'll start to see flavors within distributions tweaked at the source level for optimization. Systems made to run as a desktop will have many different default options from their server optimized counterparts.
4) Integration will force the ultimate "killer app"
I predict an open, central authentication system will take the Linux world by storm. There still isn't a Linux comparison to NDS/eDirectory or Active Directory that makes user management across the network as simple as either of the two. While eDirectory will run on Linux, there is no open standard with a GUI management tool that automates this mass management. An authentication service whose job is only to watch resources including users, devices, applications, and files doesn't exist and can't be built without serious Linux know-how. This service, which I'll casually refer to as LCAS (Linux Central Authentication System) for lack of a better term, will be as easy to establish as a new Microsoft domain's Active Directory.
LCAS will operate using completely open standards (X.500/LDAP) and will be easily ported to the BSDs and to commercial Unixes. Unlike Active Directory, LCAS services will be portable, and stored in a variety of databases, including PostgreSQL, MySQL, and even Oracle and DB2. LCAS, like Linux, will be pluggable, so that as it matures, management of other objects, like routers and switches, your firewall, and even workstations and PDAs and eventually, general network and local policies, will be controllable from your network LCAS installation. Perhaps, in time, it will also manage objects on the internet and how they can act within your network. I envision the ability to block, say, a particularly annoying application's HTTP traffic, the ability for certain users to use specified protocols, or installing internet printers via LCAS.
5) Releases will become less frequent, and updates more common
There is a competition for versioning in the Linux world, as though higher version numbers are somehow "better." Version inflation is commonplace, with companies incrementing the major version for minor overall updates, and going from X.1 to (X+1) after a few application updates and a minor kernel increase. There is also a software trend that eventually, when the version number gets too high and is abandoned in favor of less harsh sounding names. No one would upgrade Windows every six months, so why upgrade Linux every six months? Because the software gets better too quickly! And the best way to get new software that works is to upgrade the whole distro! This is backward. The software should be incidental to the distro, not the reason for its version stamp.
Gentoo Linux just changed their release engineering guide specs to include for a year number with subsequent point releases. This, I think, is the right idea. I predict that we'll start to see releases like DistroX 2004, DistroX 2005. As a counterpart, we'll begin to see downloadable updates like service packs, like DistroX 2004 Update2. These updates will be easily installable and will update and patch not only the OS, but all components that came with the distro.
It is not unlikely that we'll see a front end installer that launches, checks your system and the server, asks which pieces you want upgraded, and then processes it. There are systems like this in place today, however, they are constantly updated. Too often, people don't patch or update, they just reinstall. We're going to see only security updates for each distro, and approximately quarterly, we'll see an official Update. Updates distributed in this fashion are much more likely to be applied by a common user than the slew of updates issued on an almost daily basis. Updates like this allow users to utilize a modern system much longer in between releases - for years in some cases. Unless OpenCarpet catches on, I see a service pack mentality prevailing for all commercial distributions.
6) Linux-approved hardware will become common
Part of the fight for useable Linux is with device drivers and hardware. Getting your video card to work properly, even with a binary driver available, is still way too hard. While this isn't always the fault of the hardware, we will see, in time, Linux approved hardware. The hardware will include Linux drivers on the accompanying disk. There will be a certification process that tests hardware against a certain set of standards. Soon, a Tux badge on a PC case will be as commonplace as the "Built for Windows XX" stickers on most cases today.
I don't claim to be visionary by any means. I also don't want to forcefully bring spirituality into the mix, but I believe all things exist in waves, with highs and lows. Linux started small, it's gained an audience, and as it swells to a large point, we, the community, should anticipate the future refold of things. The eventual downswing shouldn't be an implosion, but rather, an opportunity to organize and streamline the existence of free software. It doesn't have to be a reduction in use, it can be a simple cooperation, reduction of market saturation, and convergence towards standards.
Within the next two years, we'll likely see Linux kernel 2.8, Gnome 3, and KDE 4. We'll see exciting new projects. We'll see many new Linux distributions and many existing ones disappear. We'll see the pre-Microsoft Longhorn media blitz. And I bet, not too much longer than that, we'll see some of the above start to become a reality as well.
Adam Scheinberg is a regular contributor to osnews.- "The Future of Linux, Page 1"
- "The Future of Linux, Page 2"
