A Conversation with Linus Torvalds

Now for the facts behind the news.

“Linux For Suits” was our first Linux-in-business event*, and it featured an impressive series of panels, starting with my one-on-one conversation with Linus Torvalds himself, following Linus' own opening remarks.

You can read the entire conversation below. It is transcribed verbatim. Aside from demonstrating that few people speak in final draft (and that Linus is closer to qualifying than I am), the whole thing demonstrates the extreme difference between an actual event and the news that issues from it.

On the one hand, Linus had some very intesting things to say about the nature and shape of Linux and open-source development and how that pertains to business—and also about a few other topics, such as his interest in embedded uses for Linux. On the other hand, what made the news was mostly his remarks about Sun and Microsoft (plus, to a lesser extent, his remarks about embedded Linux).

So here we have a fun chance to compare a verbatim interview with the coverage that derived from it. Here are just a few examples:

Transcripts of the other panels will also be up soon both here and at LinuxForSuits.com. And you can listen to the same events in our own Radio Free Linux series at ON24.com. In fact, ON24.com will also show the videotape, if you want to watch it via RealPlayer. Our thanks to Internet World for granting us permission to use both.

- Doc Searls

Bernd Wuebben (KDE), Michael Cowpland (Corel), Nat Friedman (GNOME),Bernie Thompson (Cosource), Linus Torvalds (Transmeta)

Doc: Welcome to Linux for Suits, the first Linux and Open Source executive forum here at Internet World. We have put together quite a day here to go over something that is a huge phenomenon just now breaking upon the business world. I'd like to ask how many people here consider themselves suits? I'm not seeing that many of us “decked out” this morning. Who admits to that? I'm just curious, how many people here had not heard of Linux two or three years ago? Quite a few. How many people had not been using the Web, say, four or five years ago? How many of us use the Web every day?

There is a pattern here, even though it isn't that obvious to see. The Web was not something we could easily predict - the Internet was not something we could easily predict - it's not something that somebody owned, which is a critical thing. Nobody promoted it, and still it is something that exploded on the world and now we are all part of it even though nobody owns it. Something very similar is happening with Linux right now, and I don't think we understand it any better. I want to get some understanding on how Linux works and what is behind this phenomenon and how in many ways it resembles the wave of adoption we've seen with the Internet over the last five years.

What we have with Linux is a radically different model for developing software than most of us are used to. We go to the store and order something that comes in a block of bits. Linux is something radically different. Linux is something where the people who build the stuff actually sign their work, so that if other people want to see who did it - they can. It isn't just a matter of wanting to get the credit or the esteem of their peers - even though that is a powerful motivator. It is also so people can find out who did this work and who to call when a problem shows up. An interesting thing yesterday: I was talking to one of our guests later on today, Jeff Davis with Amoratta Hess, who said, “You know, when I need help, sometimes I just look at who worked on this or that part of the kernel or this or that part of Red Hat or some other part of the software, and I call or write them and find out what I can do with this piece of software.” This is not something you can do with Microsoft; this is not something you can do with most software.

The person who sits at ground zero of this phenomenon is Linus Torvalds, who has joined us this morning at great expense. He just came in from Helsinki to San Francisco and flew a redeye in here to join us - I think he looks in remarkable shape considering! Linus is going to share a few words with us, then he and I are going to sit down in front and have a brief interview followed by a question and answer session. Here is Linus!

Linus: I actually had a stopover in San Francisco, and I had been home for a week before I came on the redeye to New York. I'd like to actually explain the kind of strange format here. If you read the opening remarks, you think I'm going to give a speech, and I am not. There are several reasons: one is simply, that yes, I did get the redeye, and I probably couldn't keep track of a speech for more than five minutes and then I'd just go out in strange directions because my mind wanders. The other reason is simply that I hate writing speeches. I don't have anybody that does it for me and it is just a waste of time in my opinion, so the end result is sometimes strange. But the real reason and the fundamental reason is that I thought there would be a lot more “suits” here. And there is one rule that everybody should always remember, and that is, you should never, ever let a technical guy talk to a suit. Because if you do that, the technical guy will say something embarrassing about the product or something, and it just never works out. And you have to have a chaperone, a PR person or a marketing person there, to filter the information. That's why we set up this question and answer session, and Doc is the chaperone. As it turns out, you seem to be a lot more technical group than I would have assumed from just the title of the talk. That is fine, because it probably means that if I tried to make a talk for “suits” it wouldn't have gone over very well, and this way, we can maybe modify the interview session a bit. People can probably get their own input on what they want to hear about. So I will turn it back over to the chaperone, and we'll see where this goes.

Doc: This is being videotaped, so there may be some “suits” in the overflow room, but we can keep them out of sight. The first question is a bit of a “suit” question: did you have any idea when you wrote Linux - when you just finished it for the first time - that anything like this would happen?

Linus: Well, obviously not. The whole situation has been completely unplanned and fairly chaotic. I use chaotic they way physicists or mathematicians tend to use the word “chaos”. There is an order; it's just that the system is so complex you can't see the order from any kind of small single action. It is an order that exists in the system, and that has probably been the saving side of the project. There hasn't really been much of a long-range plan; there wasn't an initial goal that is even close to where we are now. I mean, there have always been goals, but they've always been modified very extensively by new people getting in, and just when you reach a milestone, you notice you actually wanted to go in the other direction a bit. It's been a very dynamic system and the fact that it actually has worked - hasn't exploded, that it really hasn't become chaotic in the traditional sense - shows in the non-mathematical sense that there is actually more of an innate order to how large systems work.

Doc: I think what is happening is that innate order is penetrating more and more companies out there. The Linux development community is not just something that is academic or scientific, although it may also be that in a strict sense. One piece of interesting data is some people from IBM (I shouldn't characterize this as all of IBM because IBM is huge) at Linux World in August told me about how out of 600 people they surveyed for Linux awareness in one technical division, not only did all 600 know how to spell Linux, but 120 of them were hacking kernel code. What IBM is aware of and a lot of other large companies are aware of is, “This thing has infected our organization!” But it's not really an infection any more than the Internet is an infection. There was a lot of resistance to the Internet at first, too. I see less resistance to Linux in part because I think there is manifest in Linux some truths about the way development ought to be.

Linus: Well, you may be seeing less resistance to Linux now, but the situation has certainly changed. One of the things I really enjoy about having commercial companies come in - people always ask me if I mind somebody taking Linux and trying to make a buck. And the answer is, “Hell no!” I want these commercial people to come in and open doors to places Linux wouldn't have gone otherwise, because it to do the stuff I'm interested in which is the technical side and allows me to do it in places where I couldn't have done it otherwise.

Two or three years ago, there was very much a barrier, and the only way Linux tended to get into small or even large companies was through engineers, through the back door. What's obviously happened is that the kind of PR we've gotten - the Department of Justice lawsuit certainly hasn't been detrimental to Linux either - has meant suddenly you have the suits knowing about Linux, which means it's even easier for engineers to explain that they are bringing in Linux. But it is sometimes starting to happen that the suits themselves tell their engineers, “How about this Linux solution?”, even in areas where it might not make any sense at all. So it's changing.

I agree that a lot of people are starting to realize that the Linux way of doing things isn't that strange - even for a business. I've been talking about “openness” to business when I'm going to these private business talks, and almost uniformly, the businesses say, “That is what we are trying to do internally”, that is, trying to get rid of some of the fairly strict, hierarchal nature of the company. There is this open-doors policy that just about any modern company has in trying to make communications flow more reasonably across different pieces of the organization instead of always going through the channels. What Linux has done basically is exactly that, except it has taken it to the ultimate extreme, which is there is no outside to Linux - there is no external shell at all, because everything is done basically with no real channels. Everything is done by just letting information flow as freely as you can. And it works - and it has worked very well so far.

Doc: How many people are involved right now in Linux development?

Linus: Well, how many people in this room have written kernel code? About three or four. How many people in this room have written code for anything that has shown up on a Linux CD ever? So the point is that these sets were completely separate, which is strange because the kernel coders should have raised their hands for the second question. But the point is, depending on the question, the answer is different. When it comes to the kernel, a few hundred people have contributed noticeable amounts. Of those, 10-50 have contributed major chunks of code. But when you look at Linux as a system - and you really should and it's not even just Linux at that point - obviously, there are huge numbers of people. Part of being open is that it's so hard to say where the boundaries are, because a lot of the tools used under Linux are used on other operating systems too. Does that hurt Linux - no. Does that hurt the others - no. It's quite the reverse. Being open and not having a boundary between Linux and BSD in many of the tool areas makes everybody stronger. That also makes it very hard to count, whether it is 17,000 or 25,000 or something else, who knows?

Doc: I'd like to look at this word “open” for a second because some of us have been around UNIX for a long time, especially the promotional side of UNIX. I mean, Sun for years talked about being open. Somebody from HP once told me, there is nothing more closed than something that calls itself an open system. In many ways because some of these things just didn't get along; you'd compile them and they didn't agree. Now, an interesting thing - I don't know how many of us are wearing these badges with a promotion for a company called Real Names on the top of them - these guys announced yesterday that they are going open source, so we have a little promo going on right here. Open source has become something that is very advantageous, marketing-wise, to say about what you are doing. Sun just went open source with Solaris, although I might characterize that as a kind of share-cropping strategy, where you get to work on our crops as long as we get to sell them. So, I'm not sure that is as attractive as Linux. Then in many ways, the free software movement has sort of morphed into the open movement. I would take it that “open” literally means you can look at the code and then we go on from there, is that true?

Linus: Well, first let me talk about the political side. “Open” used to be a dirty word, then a year and a half ago people started trying to find a new term for what was called free software and open source was coined. A lot of people thought that was a really bad name, because of the history of misusing “open” as just a marketing term. To me open means more than just being able to look at stuff. If you were just able to look at it, it would not be open. A window is not open - sorry about the confusion, I'm not talking about Microsoft - a window is not open just because you can see through it. The definition of open is that you can enter it and start playing with it and make your own decisions. You don't have to ask for permission in order to do stuff. I think that is the important part, the fact that nobody controls what you do. Basically, all open licenses control to some degree how you can close it again. There is a lot of argument in that area where people disagree about how you can takes parts of something open and try to close it off. But the important part inall true open licenses is you can basically, at least for your own use, do whatever you want and you can also talk to other people about it and share your changes.

Doc: Would it be fair to compare it to, say, a spoken language in that respect? New words can show up, you can use them, you can contribute them?

Linus: It certainly is a language, but I don't think that analogy is very good. In fact I don't know any good analogies. The usual one that people tend to use is the hood of a car - you've all probably heard that one. Where you say that open source is like having a hood that you can open on your car so that you can tinker with your car. But most people don't want to tinker with their cars, because, quite frankly, they have more electronics than most large cities had 20 years ago. The point is that it's not that you want to tinker with your car, it's that you can take it to the shop down the street and let them tinker with your car - you don't have to take it back to Ford or Toyota or whatever. That is still a bad analogy, but it is the best one I've heard so far.

Doc: Let me try one that has been kicking around a bit, and I will admit to pushing it a bit. As a writer, I look at language and how people use it. It is interesting to me that when we talk about software, we use these construction metaphors: we build something called a site, we develop, we design, we architect, there are all these building metaphors. Inprise, formerly known as Borland, has just announced a whole raft of Linux tools they are going to develop called “something Builder”. This is interesting to me in the sense that I think what we are seeing to some degree is the software business turning from one we understand entirely in terms of its suppliers - those suppliers like Microsoft and Lotus and others that make a block of bits in a box. If you read the papers and magazines, you read about this war between this supplier and that supplier, but the truth is when it comes down to it, people build something, they need - “scratching an itch” is an expression often used about why a programmer does a kind of work. I'm wondering if that works for you?

Linus: I think that building is correct. There is the analogy where you can choose to get a standard house completely built for you: a house in which you don't have any choice. But hey, you don't have to make any decisions, and a lot of people want to avoid making decisions in life. Or you can decide to customize your house, and when you customize your house, maybe you buy one set of walls from one person and the roof from another person. That's kind of where the Linux approach comes in - that you have these building blocks to use to your own satisfaction, and you can make a house that looks different. In the end, it's cheaper; even if you buy blocks, you put them together in a very similar way. There are no surprises - you know where the closets are, and they are not in the attic. So when you go into a friend's house, you may initially be confused as to where the bathroom is, because you have made the choice of making something that fits you better. That's very powerful, especially in businesses, where you have practices that have been honed for tens or hundreds of years and suddenly you bring in computers, and you want to build the computers so that they fit your business practices or your personal tastes or whatever, instead of having to take this prefabricated thing you get from one vendor.

I think this is the future of software, simply because software is so cheap to copy. Being cheap to copy means in the long run there is not going to be much of a industry in doing standard blocks. You are going to take the standard blocks for granted whether they be an operating system, whether they be the windowing system, or whatever. What will really drive the software industry will be customized special software for special needs, and I think that is where the software industry is going. Right now, we have been in this strange situation in that the industry has been changing so quickly, even the standard blocks, that right now when we look at most software companies, where they actually make their money is the standard blocks. I think that will go away, and what people will make their money off of is the personalization. You see the same thing on the Web. That's already happened to a large degree. All the money used to be in the infrastructure in getting it working at all. Now, people are interested in the personalization - the portals - the ways for users to get their own web interface. And that is where everybody is going. I think that the Web has moved a lot faster than the software, partly because software is complex; it's a complex area, I'm certainly not saying people will change it in the next two years, but it will change in the next ten.

Doc: It is rather interesting that the majority of the Web pages we see today are served up by open-source software - by Apache often running on Linux, Free BSD or one of the others. This is software that was developed on this model - it wasn't developed on the block of bits model.

Linus: Well, it's partly because of the flexibility. Obviously, when you are doing web stuff, especially three, four or five years ago, the whole market was changing very rapidly and you needed the flexibility. You could not go in many cases with a completely finished solution. In many cases, a lot of the ISPs basically had to build up their own solutions. That's where the prefabricated building blocks really help. You can basically do whatever you want. Then what happens, is that after a time, you reach a kind of stable platform, and the building blocks are bigger. The building blocks that used to be individual bricks are now these huge walls that you just put down next to each other when you build a house. The same thing happens in software. The building blocks five years ago used to be the OS kernel and the compiler and whatever, and these days the building blocks to a large degree are more of a combination of all these so that now you have a building block that is a base system, windowing or web serving, and you build your own house that way. So change is happening in many ways, but the same fundamental building-block analogy does hold true.

Doc: I'd like to invite people to come up now and form a line at the microphone.

Simon Applebaum: I'm Simon Applebaum, one of the senior editors with Cablevision Magazine. Linus, the company you are a part of, Transmeta, is run by Paul Allen, and Paul Allen runs Charter Communications, a major cable operator. Has Paul Allen approached you, or have you approached Paul, about using Linux as an operating system for the cable industry through Charter's digital set-top boxes? And if not, what role do you see Linux playing in cable as a set-top box operating system?

Linus: Okay, first off, I can't really talk about Transmeta. But I guess I can say this much - no, it is not run by Paul Allen; it's run by Dave Ditsell. Paul Allen happens to be one of the people or one of the venture capitalists who put in some money. That's all of the connection there is, really. As a result, no, I have not talked to Paul Allen and he hasn't talked to me. I've talked to other people interested in embedded systems. For some reason, within the last year or so, the embedded market has sprung up as something really interesting. It's not something I would be working on personally, but it's something I've been getting in contact with through a lot of different sources, and then set-top boxes have been the obvious kind of high-end embedded system people are looking at. I don't know how many of you know Tivo, the digital recorder, that's a kind of set-top box running Linux, even though they don't actually mention it in their ad.

Simon Applebaum: By the way, in a story we did a few weeks ago, we mention Tivo is running Linux and Tivo is also going to be a cable player as well.

Linus: The fact is, most of these embedded people do not even really talk to me - and I consider that a feature! When I'm talking openness, I really mean it. It doesn't mean I want to control things. I want to be out of the loop, especially on issues that aren't my specialty. And cable TV is not my specialty! So, I get a lot of technical questions and usually they are from people I can't talk about openly. And there is a lot of interest in this area. But what will actually happen is like with the Tivo: I will actually be told two months after the announcement that, oh by the way, it's using Linux! So, I don't know any more than you do.

Amy Wohl: Hi, my name is Amy, I am an industry analyst. I'd like to ask you a question about the very interesting point of view you expressed about going to a world in which software is built out of building blocks and is therefore more custom rather than the kind of shrink wrap packages we see today. That's almost a contrarian expression. Yesterday, Larry Ellison in talking about Oracle said he wasn't going to allow anyone to change even a single line of code in Oracle applications: certainly, a strict point of view. Do you think when this building-block approach begins to get implemented, it will be through people actually doing coding or do you have in mind more of a model in which people are able to put the blocks together the way they do personalization on the Web today by making choices through some sort of dialog interface?

Linus: I think the coding part is obviously very important - that's the part I've been involved with. I think, in the long run, most of the customization work is not going to be so much coding, as it is going to be designing a system. For all I know, Larry Ellison might well be right in that nobody is ever going to change Oracle. Because Oracle will be just a small building block; I mean, it's considered large, but if you actually think of it as a part of the whole infrastructure at a company, it's a small part - it's important, but it's still just a building block. You will see a lot of these kinds of customization packages. You already see them. This is why Visual Basic was so hugely successful - not because it was a good medium for programming, because it's not. Don't get me wrong, I like Visual Basic, really, and I think it is one of the best successes Microsoft has had. That was because it was a really good medium for customization, it was a great medium to the front ends for the real work. I think that is the future.

Audience member: I have a bit of a different question. Mr. Torvalds, your home country is Finland, which has been receiving a lot of press lately - Nokia's success in giving wireless to the world, a major investment by a Finnish telecom operator in Boystree recently, the highest Internet penetration in the world, the highest cellular penetration in the world. What is it about Finland?

Linus: Finland is an absolutely great country, except for the weather! The weather sucks. Everything else works really well. I think that one of the issues with Finland, to just kind of bring it back on track with Linux, is that it's a fairly small country and is very homogeneous. It's kind of a building block. If you look at cell phones, it's extremely hard to introduce cell phones in the U.S. because the U.S. is a very, very complex economy, very diverse. There is a lot of existing old infrastructure and it is way too expensive to upgrade in one go. And this is kind of similar to Linux, where a small and agile person, or a small and agile country can actually do really well. In Finland, for some strange reason after World War II, the technology side was emphasized very heavily, and because it was small and easily accessible, it was very easy to upgrade.

I moved from Helsinki to the Silicon Valley, and boy, Silicon Valley is supposed to be the center of the universe when it comes to technology, but it's a third-world country in many ways! A lot of the infrastructure works so badly. Stuff like electronic banking is electronic in name only; the banks actually send checks on paper between themselves after you use your web browser to fill in the information. That doesn't happen in Finland. When you push the send button, it is on the other person's account - instantaneously. The other reason, apart from these benefits of being a small, agile country, is that the weather does suck - the winters are long and dark, and there is nothing else to do.

Doc: Do we have any technical questions?

Audience member: Do you see the knowing interface, the user interface, completely diverging from the operating system, the underlying system?

Linus: To some degree, yes. The reason again is that in this situation you don't want to make your systems too complex. The way to avoid complexity is by having regularity and flexibility, so that people can pretty much mix and match. One of the issues, for example, is that Linux or any operating system is supposed to be usable on a lot of machines where a windowing interface doesn't even make sense. This is something that Microsoft has several times found out the hard way. First with Windows 98, I noticed that it doesn't work that well on servers. Then with Windows NT, I noticed it doesn't work that well on most embedded applications. Putting the windowing interface into Windows NT, a system where it doesn't make sense to have it, was stupid. You don't want to have that type of coupling. I'm sorry for picking on Microsoft, because others do the same stupid mistake over and over again. It's just that some mistakes are more obvious. So what you really want to do is decouple as much as possible, because it gives the user the flexibility to use just one or the other part but also, from a purely developmental standpoint, it means you have a chance of actually understanding how the system works technically. For example, if we tried to integrate the X server into the kernel, or even worse, if we tried to then integrate the windowing system, the end results would not be pretty. There are UNIX vendors who have tried this and they've always failed, so they've put in resources. The reason is simple: you want to go for regularity and you want to go for interchangeable parts.

Same audience member: Then as a followup, how would you feel about having a Microsoft Windows lookalike skin or interface that would actually help you gain a much bigger desktop space, which would have a much more reliable underlying system?

Linus: I would probably not use it myself, but I certainly would not mind.

Same audience member: Do you know of any such projects going on?

Different audience member: Well, there are skins for Enlightenment that will allow you to do that, as well as for Window Maker and I think FVWM95. They are all very familiar to Windows users.

Linus: There are interfaces that make Linux applications look more like Windows applications. If you want to go on a deeper level where you actually want Windows programs to not know they are running on top of Linux, there are solutions for that too. But you get into a much hairier country.

1st Audience member on this topic: It's just that the issue is the migration, a lot of corporations - a lot of big companies - would love to switch over to Linux, but the serious issue is retraining all those users who do nothing but use a basic word processor.

Linus: Yes, for retraining purposes, it's usually the windowing interests that seem to be converging on something that looks similar to the Mac or Windows or something like that, just because people are used to that, and what people are used to, they like. For migration purposes, one of the big issues is often that there are like one or two applications that you want to run and that iswhere solutions like Soft PC have come in.

Audience Member: I think that Microsoft's IE5 is technically a lot better than Netscape if you just think about XML or however - how do you think Netscape or an open-source group can catch up to Internet Explorer?

Linus: Well, I'm not going to go into Netscape, but it does bring up a separate issue that tends to come up over and over and over again which is that people think open source automatically means that the end result is better. That is not true. Open source means you have the potential to take advantage of a much more scalable way of doing development. The fact that you have the potential to do that does not mean that you will, or even if you will that you will actually lead in the right directions so that it succeeds. It is fairly obvious that Netscape, as an example, did not get the interaction going between the outside world and the Mozilla project itself. Which meant that, yes, it became open source but to some degree, it never took off as they hoped. The good news is that because people are now thinking Mozilla didn't work, suddenly outside groups are feeling more empowered. They think that okay, Mozilla failed, let's take the Mozilla source now and make our own modifications. There seem to be a lot more groups that are more interested in pushing the browser issue. And Mozilla seems to be less arrogant as an organization. So, I actually think that this situation may change. But right now, this is an example of no, open source is not the panacea; the answer to world hunger or whatever!

same guy: Yeah, but actually the Internet is expanding. Everyone will be using the Internet in some years. However, there are already some sites that are not usable with Netscape that I have experienced. What are you doing so that people can use the Internet with Linux too?

Linus: Me personally? Nothing! I'm making sure that the PCP stack works fine! But that is a slightly different level than Mozilla! What I feel is that people are worried about playing catch up with changes. That's only true as long as web content is actually dynamically changing all that much. It's not anymore. Yes, right now, for example, IE5 shows some pictures better, some sites show up the way the author intended for them to be shown up. But the changes aren't that large. How many of you use Netscape? How many of you have a problem? How many of you blame the site? There are not that many real problems. And the other thing is that yes, it actually is the site who loses the customers. Which is not a problem because a lot of these IE5 only sites are Microsoft-only sites. So, you look for your customers and you go for your customer base. But if you look at anything that even wants to be big and important, they make sure that their site works with Netscape too.

New question: Microsoft has been talking a lot about opening up their own code. Whether or not they actually do that is a big question; but if they should do that, what does it mean for Linux?

Linus: I'd like to start off by saying talk is cheap, and I think this is a fairly theoretical question. However, as a theoretical question, I'd be more than happy to see more and more people open up their source. I'm not convinced, for example, that the Sun community license is a very good license. But it still makes me very happy to see that Sun is opening up and making their knowledge available to others. And if Microsoft were to open up, I'd be more than thrilled. I don't think it is very likely, or if it does happen, it will more likely be in niche markets. People think there is more of a Microsoft bigotry in the Linux community than there really is. I'm more than happy to use Microsoft products. It is just that I am selective about the products I want to use. For example, I've always liked PowerPoint, and I've always thought that Visual Basic was a good product. It's just been hard to use them because I've always thought the platforms they run on aren't good enough. I used to do my slides with PowerPoint for the longest time with a Windows Installation package. This is true, I find of all the developers I'm in contact with, that maybe we take up Microsoft as a bad example in specific areas, but we're not as anti-Microsoft as the press makes us seem.

* “Linux For Suits” will also be the title of my new column, starting with the December issue of Linux Journal, when Peter Salus takes over my old column title, “Penguin's Progress.”