HP decided that being in the microprocessor business wouldn't be economically feasible in the long term. Why is it economically feasible for Sun in the long term?
Because I don't make the chips. It's a fabless model; I design the chips. The question is, Is there value that you get out of being able to control microprocessor design? So far, we've proven yes, there's tremendous value. In fact, as we go forward, the case gets stronger because the nature of what you want to do on processors on the server side is really changing. We're building, essentially, entire servers on chips.
HP made a decision that it did not have sufficient volume in its processors to justify it continuing to design and manufacture its own processors. So it decided to get out of the semiconductor business and get out of the processor design business, teach Intel what it knew about processor design and go create Itanium. Itanium is sort of a zero-volume processor right now, so the economics of that is a disaster. I don't understand how that is somehow a more viable economic model than what we do.
The traditional argument is that going with the Intel architecture offers economies of scale that you can't match.
Let's talk about the operating systems and then talk about processors. At the programming level, there's no difference between Linux and Unix - Linux is a clone of Unix. At the level where programmers are concerned, they're essentially identical. At the implementation level, they're quite different - Linux being built more in the Berkeley style of kernel, and Solaris following more of the second-generation AT&T kernel design. Linux is very well tuned for Intel processors, so it's a good performance equation there. What a lot of people have been after is getting access to very effective 32-bit computing technology from Intel. Where Intel certainly made its mark in the server space has been with Xeon, being able to do one-, two- and emerging four-way systems at very high performance and very good prices. The frenzy with Linux has been about really getting access to doing 32-bit computing with Intel technology on something other than Microsoft operating environments. You can get access to that with Solaris on x86, too.
Sun has already done a lot of work to port Solaris to Itanium. Was all of that work done in vain?
What's happened is that as we've moved forward to the new software with Itanium and UltraSparc 5, we all started these things before we even understood anything about Web applications, let alone Web services. So they were designed in an environment of, "How do I run a single job fast?" And that's not what we want. The Itanium is off in a corner of the computing space where I don't think the volume server is going to be. We've gone back and said it's going to be far more important to support 32-bit.
Can you foresee any scenario in which those Solaris-on-Itanium efforts would resume?
Yeah, I could see that if Itanium starts to build some substantial volume in the market. It hasn't proven to do that. I don't have anything against it. From an engineering point of view, it's real bravado on Intel's part - they're doing a very complex design. But I just don't think that's the centre of what the applications need.
It's pretty clear that there's an ongoing debate within Sun about how aggressive you should be in adopting Linux as more of a core part of your operating system strategy. Can you give me a behind-the-scenes glimpse into how that debate is going?
The debate is around how much longer people are going to persist in focusing on operating systems. It's not where people are writing applications. If you go into an enterprise, if developers are writing an application to an operating system, you really have to question what they're doing. Because most new applications are being written at the next layer up - they're written as server pages, as directory entries, as database tables, as JavaBeans. If you're writing C and C++ applications still, you're doing something old or something that's very special. But it's not where the mainstream development is, because it's simply not productive enough. In that context, we view Linux and Solaris and Intel architecture and Sparc architecture as all components and choices. You use the best component for the job at the time. Honestly, I don't care. It's not important.
s there anything to keep Linux from developing into a head-to-head competitor against Solaris?
I think Linux is an alternative to Solaris today, and I don't view it as head-to-head competition.
No, the question is whether there's anything to keep it from developing into that in the future?
As an operating system kernel, it's a component choice. The question is really about how this evolves to the next layer, and, of course, we think that's all around Java.
What are the prospects for Sun to adopt a Linux-on-Sparc strategy?
It doesn't make a lot of sense right now because Linux is a binary architecture. It's exactly the same [reason] why Linux on Itanium doesn't make sense, because the volume is around the 32-bit binaries.
For the Unix market as a whole, do you see the curve going up, going down or staying flat?
I'm not one who looks at the whole market - I look at our piece of the market. We are blending Unix and Linux systems, and I think that total market grows for us.
What do you make of what SCO is doing, and how do you expect it to impact the Unix/Linux market?
I think it makes everyone take a pause and want to ask questions about the pedigree of the intellectual property they have. When you buy a Linux support package from Red Hat, you do not get indemnified as a customer for IP violations in that code - it's just not part of the contract. If you buy Solaris from Sun or Windows from Microsoft, in the purchase contract is an indemnification. If there is some misappropriation of intellectual property in Solaris, let's say, then we are liable for that. We will stand behind it; we will indemnify our customers against being prosecuted on that. Red Hat won't do that for its customers because they clearly can't.
What do you consider to be the biggest competitive threat to Sun at this point?
Our biggest competitive threat is how long the economy stays this way. The market reality these days is that people are much more concerned with the short-term financial mechanisms that they have around IT - witness a lot of the outsourcing movements. For a lot of people, it's sort of mortgaging its future. So our biggest competitive threat comes from something like an IBM outsourcing contract, which comes into a shop and says, "We're going to give you some positive cash flow for the next year or two." Of course after that time, you're going to go negative, and it's going to be expensive for you. But this gives people some short-term financial relief. So we tend to compete with something that looks more like a bank than a technology innovator.