Andrew Morton, Linus Torvalds right-hand man and the maintainer of the 2.6 kernel, will not be attending the LinuxWorld conference in New York this week, but he did speak with Senior Editor Peter Galli about the recent 2.6 kernel release and his vision for Linux.
What will the buzz be around Linux in 2004, from your perspective?
It will be around the desktop. We are very interested in that. The Sun [Microsystems Inc.] desktop [Java Desktop System] makes sense and is a good innovation. The OSDL [Open Source Development Lab] is also doing work with some of our partners around the desktop. Novell [Inc.] also now has Ximian, so that is all looking good and promising. There are also a lot of desktop applications that do not require the whole mobile professional business suite. Things like kiosk applications, point-of-sale applications and more embedded applications. I think thats another entry point for Linux on the desktop.
Some Linux watchers have said 2004 will not be the year of the Linux desktop. Obviously you disagree?
I think a lot of mobile professionals go by their own personal experience and, frankly, trying to get all your e-mails and other things on your BlackBerry with Linux really sucked. But thats not the whole desktop market. Theres a percentage out there that dont need a Web browser or PowerPoint and all those things, and thats an entry point for the Linux desktop.
As you and Linus move toward the next test and development kernel, 2.7, are features for the desktop and desktop functionality a high priority, or are you continuing to focus on the server?
A lot of work went into the server side in the 2.6 kernel, mainly because that was the area where more work needed to be done in terms of some scalability and hot-plugability. But were all very sympathetic to the desktop as everybody in the kernel world uses Linux as their desktop, and we want to see it do well. What we did give the desktop in 2.6 were improvements in the areas of device management and hot plugability. The 2.4 kernel was not bad in that space either. So, in my opinion, not a lot needed to be done to the kernel for desktops in 2.6 since the 2.4 kernel is fine for desktops as long as we get the applications story sorted out.
What do you think the greatest changes are that the 2.6 kernel will bring users when it ships in commercial distributions? What are the parts you are most proud of?
It would be the improved scalability on the high end for certain machines and the improved responsiveness and interactivity on desktop machines.
When can we expect to see the release of the 2.6.2 update?
We just got through 2.6.1, which had a huge backlog of stuff that had been saved up during the 2.6.0 freeze period. I seem to have another 300 changes here ready to go, so 2.6.2 will also be a significant merge. You can expect that in a couple of weeks. But we are being pretty successful these days in keeping the kernel stable as we make large changes to it. Thats not something we were very good at across the 2.3 and 2.4 series. I think weve sorted out some processes a lot more now in that area.
Next page: Looking toward Kernel 2.7.
Page Two
Looking toward 2.7, is there still not a lot of desktop work that needs to be done to the kernel or are you happy with where it is now?
There have been several attempts to look at what the 2.7 feature set will look like, but I havent seen much concrete as yet. The kernel summit last July was supposed to be about 2.7 kernel features, and all that was talked about was 2.6. Nothing in my opinion has really solidified there, though there has been talk about allowing the kernel to use larger software page sizes, to get a clustering file system in there, additional clustering support. It wouldnt surprise me if those things started happening.
Some kernel developers have said they think things like Non-uniform Memory Access [NUMA] as well as an Enterprise Volume Management System need to be addressed in 2.7. Do you think this will happen?
We have a NUMA implementation in 2.6 and, while its probably not the worlds greatest NUMA implementation, its there and real and people are using it.
When do you think that you and Linus and the kernel developers are actually going to sit down and look at what should be in the 2.7 feature set?
Id expect the Linux kernel summit this July to spend a lot of time looking at the 2.7 development kernel and the 3.0 kernel.
Microsoft [Corp.] continues to maintain that it has less security and vulnerabilities than Linux. Do you think this is a fair assessment?
Bugs in the actual Linux kernel are fairly rare, even though we have had a couple of kernel bugs lately. But how that compares to the Windows kernel, I dont really know. If they say that there are fewer holes in their kernel, that wouldnt surprise me as most of the vulnerabilities are at the application level. More people are probing Windows, I think, as once you get an exploit, there are more machines it can take. But Ive also heard people argue, and Im inclined to agree with it, that Microsoft made some big mistakes 10 years ago with the design of ActiveX and their whole component model. This basically left them open to exploits when people became more security conscious over the last few years. Microsoft has a whole bundle of applications whose interfaces just werent designed for security. Im guessing that closing those holes now would break existing applications.
Are we going to see a lot more aggressive security work in the kernel moving forward?
Well, the 2.6 kernel was mostly a secure, enhanced Linux which the NSA and others worked on. Its a more sophisticated, more finer-grained security model than the traditional Unix ones. But I dont know if any of the mainstream developers have any plans to go beyond that in 2.7.