News Stay informed about the latest enterprise technology news and product updates.

Torvalds: Test kernel crucial for proactive enterprises

In this interview, Linus Torvalds talks up the test version of the 2.6 Linux kernel, which was released last weekend. He also hints at when a stable, production-ready 2.6 might be released.

Linux kernel creator Linus Torvalds last weekend released a test version of the Linux 2.6 kernel called test9, a sure sign that a production version of the next kernel is fast approaching. Torvalds and 2.6 kernel maintainer Andrew Morton hope enterprise IT shops will download, install and test the kernel, then report back issues in a timely fashion, so that any problems can be resolved before the release of a stable, production-ready version. In this e-mail exchange with, Torvalds explains what kind of insight he hopes to gain from enterprises that install test9, and he reveals a tentative release date for the kernel.

Test9 is an opportunity to gather feedback from the user community on the 2.6 kernel. Are you anxious to hear from users on any particular features of the 2.6 kernel? Any anxieties about a particular feature or enhancement?

Linus Torvalds: No, there's no particular feature that I feel extra anxious over, but the wide testing is important, partly because different people have different usage patterns, and those patterns trigger things that could otherwise have fallen through the cracks. This is particularly true with regard to hardware: getting a wide and varied test base to make sure updated drivers work reliably in all configurations.

With so much enterprise attention focused on 2.6, how is this beta and the ensuing feedback going to differ from the feedback you got on earlier versions of the kernel, given that much of it is going to come from those not necessarily involved in the open source development community? How valuable will it be?

Torvalds: What I'm really hoping for is that enterprise users who know that they will eventually switch over to 2.6 will just decide to do internal pilot projects in-house. They have the incentive to do that anyway, but, at the same time, these are also usually the same people who do not want to just test any random development kernel. This is what the test-series is all about, especially the current crop: letting people like that know that [this is] no longer any random kernel.

What kind of feedback are you hoping to hear, from their perspective?

Of course, the feedback I hope for is a lot of people telling me that they tested it and it worked so beautifully for them that they decided to upgrade their whole setup, but somehow that never seems to happen. So realistically, the more people who try out the beta kernels, the better insight we get into what has gone right and what has gone wrong.

What features or enhancements to the 2.6 kernel would drive Linux deeper into the enterprise data center?

Torvalds: Part of it is definitely improved capabilities at the high end, particularly with 16- and 32-CPU machines (and even up). And the thing about high-end support is that, while not everybody actually necessarily even wants it, just the fact that people realize that they can go out and buy it and that Linux works fine on it makes users all that much happier about their more common four-way and eight-way machines. They know it's not bleeding edge. (Bleeding edge may be sexy, but most people are actually happier if they feel that they don't have to go there.)

At the same time, my particular area of interest is actually the desktop. That's where I personally use Linux myself, and I do believe that the desktop is actually starting to become important in the enterprise (though not the data center you mention). And we've spent a lot of effort on that side, too.

How crucial is it for enterprises to download and test 2.6 now?

Torvalds: I certainly realize that not everybody will feel that they have the resources or the need to do so -- a lot of enterprise customers are perfectly happy to wait for the Linux vendors to validate the kernel for them. And, for some, that really is the best approach.

But at the same time, any IT organization that doesn't try to be proactive about their upgrades and try to find out what particular issues they have is really taking a lot on trust. And being involved early is not just good for making sure that potential bugs that affect you are found early, but it also ends up being the best way to make sure that your concerns are heard in development.

Anyway, I'm waffling. This is a decision that the IT center needs to make on its own. I can really only do the 'stable beta' releases and let people know that these things are available so that a good IT department can start to get ready.

What does the timetable look like before a production version of 2.6 is available? What happens when you collate the feedback you get? How long would it take you to evaluate things and work toward a final version?

Torvalds: Well, first off, even 2.6.0 is not really 'final,' in the sense that, clearly, the stable kernel will continue to be maintained and improved for several years to come. There are still people out there that use kernels based on Linux 2.2, and people who maintain it.

That said, the real release is always a big issue, and in the end the final decision ends up being mostly about 'gut feel.' That ends up being based on several things: obviously, on the kinds of bug reports that we see -- are they showstoppers or just annoyances? -- but also on a lot of input from the different developers.

And it's a balancing act. At any point in time, delaying the release may mean that we find and fix a bug before releasing it, but it also delays more users actually getting on board the new kernel, and that will delay a lot of good information about the behavior of the kernel.

So with the above in mind, right now the tentative schedule is to release test10 in another week and let that just simmer for a while. If that looks like it might be 'the thing,' we'll end up calling it 2.6.0 (trying hard to avoid last-minute fixes). If we find any issues that need attention, we'll cut a test11 and so on, but the hope really is that we'll be done by early December.

Famous last words.

FEEDBACK: How likely is your enterprise to download and test the 2.6 kernel?
Send your feedback to the news team.

Dig Deeper on Linux servers

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.