Torvalds Looks Ahead

 
 
By Peter Galli  |  Posted 2002-01-22
 
 
 

As the Linux community congregates in New York for the LinuxWorld Conference and Expo, the center of attention, as always, will be Linux creator Linus Torvalds. A little more than a year ago Torvalds released the 2.4 kernel and has spent much of the year working on numerous 2.4.x versions to further stabilize and strengthen Linux. After handing the 2.4 kernel over to Marcelo Tosatti to maintain late last year, he turned his attention to the 2.5 development tree. Torvalds took time out to exchange e-mail with eWEEK Senior Editor Peter Galli about his work on and vision for 2.5.

eWEEK: You have now handed over the 2.4 tree and are concentrating your efforts on 2.5. What areas of the operating system are you most concentrating on improving/creating at this early stage, and what major features do you hope to ultimately bring to the next version of the kernel?

Torvalds: Right now most of the work by far is in various device infrastructure areas, notably the block device layer is being cleaned up and streamlined (and for IDE devices for examples the code to support full 48-bit addressing for really big disks, etc., is being merged as I write this). Also, USB 2.0 is being developed in the 2.5.x tree, and will be backported to 2.4.x as the need arises. Finally, there is the generic device management issue with power-down, etc. The patches already exist for it, and those will also be merged in (and fleshed out) fairly early in 2.5.x. At the same time, apart from device issues, a new scalable scheduler setup got merged already, which does much better especially on many CPUs (that is 8+), but also scales better on load. The issues of NUMA and clustering are still open, and well see how far that gets during 2.5.x. But that is largely dependent on how much hardware and other interest there are going to be.

eWEEK: Now that a year has passed since the 2.4 kernel was released, what are you hearing from developers and enterprise users about the features it brought? Have you seen a spike in interest from corporate and enterprise users or a surge in interest for features that directly address their needs?

Torvalds: 2.4.x definitely addressed the scalability complaints pretty much across the board, and you can see that in how people use it. Well go a bit further along that way still, of course, but I think we basically solved the issues that a number of enterprise users had.

eWEEK: What, if any, aspects of the next kernel will address the needs of large, enterprise users, particularly now that you have come so far on the scalability front?

Torvalds: There are two parts to this. One is just the fact that the scalability tends to mature a bit—the same way 2.0 introduced SMP and 2.2 made it mature, the scalability of 2.4 will mature over the 2.5.x development timeframe. The scheduler work that has already been done in 2.5.x is an example of this, but well have other cases where we can just improve on the existing infrastructure. The other part is the less certain issue of clustering—we will see what part of that effort makes it in during the 2.5.x timeframe, and that will in fact be one of the deciding issues for whether the next version will be called 3.0 (clustering) or 2.6 (more incremental changes).

eWEEK: Do you have a provisional timeframe for the release of the next version of the kernel?

Torvalds: Not really. It depends a lot on how I "feel" about the progress in another few months.

eWEEK: What are your goals for Linux in 2002 and where do you think its going?

Torvalds: I dont have any huge goals for the kernel proper. A lot of what the 2.5.x work has been is actually largely cleanup and supportability compared to many of the quite fundamental things that happened during the transition from 2.2 to 2.4. Which is just good. Ive told people for years that most of the new cool stuff is in user space. All the GUI work, all the cool hacks, etc.

eWEEK: Is Linux adoption negatively or positively affected by the current economic climate? Some people say the fact that it is an affordable OS drives adoption in tough times, but critics like Microsoft say that it is just as expensive to support and so is no more affordable. What are your thoughts in this regard?

Torvalds: I just dont think economic forces are all that important at that level. I dont think the people who adopted Linux a few years ago did it for any "boom" reasons, and I dont think the people who adopt Linux today do it for any real bear reasons. It makes a good story how Linux saved Amazon millions, but I bet that was a after-the-fact big perk, and that they started using Linux largely for other reasons. Dont get me wrong, I think economy matters, but I just dont think it is driving [Linux sales].

eWEEK: What are your thoughts on the talk that AOL is looking to buy Red Hat and, even if it doesnt buy them, wants a closer relationship between the AOL client and Linux. Whatever the outcome of the talks, do you think this will be good news for Linux on the desktop and its more

Torvalds: I have a really hard time judging that, especially since the rumors dont seem to be that strong yet. And in any case the "marketplace" stuff is just not something I try to get involved with.

Rocket Fuel