This year will see Linux finally crack the lucrative desktop market as more commercial software vendors tool up and cash in on the operating system and kernel developers improve graphical interface integration, says cult hero and Linux founder Linus Torvalds.
Computerworld: How do you feel Linux on the desktop is progressing?
Linus Torvalds: Last year was good but I'm seeing a lot more noise about it this year. The server space is easier to tackle first with any operating system as it can be applied to specific tasks such as mail serving; however, the desktop is harder to sell.
Now, the kernel and other pieces are coming together including office applications, games and Web browsers. This has made the Linux desktop interesting to commercials. Commercials tend to choose one desktop, such as KDE or GNOME (GNU Network Object Model Environment), and stick with it. There has been some confusion and rivalry that has helped its development. Right now it looks like the two are closing in on each other, for example, with Red Hat's Bluecurve interface.
I don't think X is going away as it has a powerful infrastructure and throwing it away would be stupid. And its network transparency is good. It's likely that X will be the 2D interface to a lower-level graphics system that is based on OpenGL. The Linux desktop wants to have 3D as the base and X as the interface to 2D.
The fact that X and kernel development have been separate is good; one could evolve without the other but DRI (Direct Rendering Infrastructure) has made them not completely independent. As a developer, having the two separate is good, because different people are good at developing for each.
Any plans for 2004?
I've never had much of a plan for the direction of Linux as I react to outside pressure. This year there will be a lot of desktop users, which will impact kernel developers.
For now, I will be working on the stabilisation of kernel 2.6 and in a month or two I expect Fedora (the core of Red Hat Linux) to have a release with 2.6 so I expect to get more bug reports.
Would adopting an integrated hardware and software system be good for Linux?
There are pure technical disadvantages of having an operating system that supports a wide range of hardware. The variety of hardware makes it challenging as Linux needs thousands of drivers.
But having an operating system that is independent of the hardware is powerful for the user, as it is basically the same on big and small machines. Another big advantage of a wide hardware base is an operating system that is more flexible. This is why Linux is having a lot of impact in the embedded space. An operating system is a complex beast, so it's nice to have an existing one that can be adapted to the hardware. There are a few problem spots with Linux driver support by hardware companies and wireless is one of them. With hardware getting better this problem is being solved.
What about Linux in the enterprise?
The direction Linux takes in the enterprise will depend on what resources enterprise companies put into it. This is the one thing that will push Linux into the high end.
IBM is the most obvious, and although it is impressive to run Linux on high-end hardware, most of the people who work on Linux don't have access to it. It's the regular desktops that get most of the attention by programmers.
What about open source code bundling by commercial software companies?
Quite often that's fine and it is fine with BSD (Berkeley Software/Standard Distribution) code. But I like the GPL (General Public License), because I want people to give (code) back. If hardware appliance makers don't give back code then that's a problem, but giving it back shouldn't cause any problems. And quite often they don't need to modify the kernel.
How do you view the relationship between free and commercial software?
Software is not becoming free but it is becoming a commodity. Once you have a commodity product the things you make money on are the services and hardware that are built around it. For example, with a lot of mobile phones the software is not the value in the product.
Shrink-wrap software businesses are the exception, not the rule. There are very few pure software houses as it is usually sold as a package.
On the desktop, it's hard to say what the commercial applications market will be like because it hasn't really taken off yet. On the server there are already companies, like Oracle, that don't have any problems selling commercial applications.
Open source is good for general software such as the kernel and development tools, and commercial software is good for specialist software.
For example, MySQL is doing well in the general Web serving space, whereas Oracle is doing well as a specialist business database.
What are your thoughts on the recent SCO issue?
This week has been good and I'm happy to see Novell release a letter that SCO is violating Novell's agreements. SCO also had to make available its case to IBM. This reaffirms the fact that this is not about copyrights but a contract agreement with IBM.
It's been very irritating at times with SCO's ludicrous, unsubstantiated claims. Some of the press has picked up the SCO case without a lot of critical analysis but lately SCO press releases have (been subjected to) a lot more scrutiny. Outside the U.S. SCO has not been good at pushing its case. I don't have a PR department so unless journalists come to me I have no way of (commenting on) SCO.
Lawsuits are a big part of the business landscape in the US. It's good that this case has made all the Linux developers aware of code, but it has been bad because it is irritating and I definitely don't want something like this to happen again.
All the Linux developers take copyright very seriously. They are developers and want to do coding, not copying. Because of this, I feel that the code quality of Linux is even better than commercial Unix operating systems. I'm not worried about copyrights but the Linux community doesn't have a lot of lawyers, PR or marketing.
What has working at the OSDL (Open Source Development Labs) been like?
Working at the OSDL has been good and although it's based in Portland, Oregon, I work from home in San Jose.
A lot of the vendors that make up the OSDL want to sell hardware and support services so I don't see any conflict with joint Linux development between them.