Doug Engelbart's    Colloquium at Stanford

An In-Depth Look at "The Unfinished Revolution"

Thank you, Doug. Before I get started I'm going to be talking about open-source. I should mention I'm not the worlds expert on open-source and I know there are many of you in the audience today who are quite knowledgeable on open-source. So, if I say anything that needs correction please jump right in, don't wait, we don't want to confuse anybody in the audience. Before I get started, how many of you here today already are fairly familiar with the concept of open-source? How many of you already feel that the open hyperdocument system really needs to be open-source? Okay, I guess I'm done and I'll just sit down now. For those of you out in the electronic audience, we had a very high show of hands on both of those questions. I'm not surprised. The open-source community for those of you who may not be familiar with it, I think is the best example of a networked improvement community that I can think of. It's really remarkably well organized considering that the organization comes from the bottom up. In other words, there is no top down organization, there are no governmental or corporate structures that make the open-source community work. The definition of open-source, that's on the web and I'll refer you to it at opensource.org, there is more to it then just the source code of software needs to be available. In addition there are other requirements, free redistribution without royalties, you have to be able to allow changes to the code and no discrimination. For example, it might be tempting to say, we want our software to be completely open-source and available to everyone, except for people at this one company. Now I won't give any names of a particular company, you make up your own company. The fact is you can't do that. It is either open-source to everyone or it's not open-source to anyone. That's why Sun, which has a real issue with one particular company, doesn't feel comfortable with the open-source model. They have come up with their own model, which we should cover; I think it's an interesting model. Examples of open-source software include Linux, of course, the operating system. Even more interesting then Linux, I think, is Apache the web server, which is extremely successful in the market place. I believe its still gaining market share, I think they have the highest market share and are still gaining market share against Microsoft for example. The open-source model is starting to catch on among many of the companies; in fact Apple is going partly open-source with their new operating system. HP is doing some open-source stuff, of course SGI has gone totally over to the open-source model as an attempt to come roaring back as they used to be. The only one who's really not playing, well Microsoft is not playing; the only one that you might expect to play who is not playing is Sun. They have gone for a different model; they are going for the Sun community license. The reason for that is Sun gets to keep ultimate control of the code that way and they feel strongly and sincerely that this is the way to go and it will benefit everyone. It's an experiment that I think we should have some sympathy with. I don't think we can say we've explored license space thoroughly for software, its early days yet. So I think if Bill Joy wants to do some experimentation, I think surely he's earned the right to do that given what he's given to the community. 

So what is required to have this process work? It's been said, by the fellow who did Tickle John Austerhout, that it takes about five thousand users to get a really healthy open-source effort going, that's users not developers. In other words, the pressure from five thousand users is enough to get a robust process working. So that will be our goal, I feel, that will be my goal for the open hyperdocument system, is to get whatever we come up with good enough so that there are five thousand people using it in a real way and then we'll have something that will successfully boot strap, continue to boot strap, we will have succeeded I feel if we can get that far. I should mention that open-source software is not a new idea at all; you know the previous term was free software and even before the term free software original software was all open-source, it was traditional for software to be open-source from the beginning. There has been no invention of an extremely new theory of software, in a way we have gone back to the roots of what worked in the early days when the Internet was originally developed. The reason that the name changed from free software, which is still a viable name and we use that sometimes when we want specify that specific subset, the reason it changed from free software to open-source is primarily because it was very hard to get companies to pay for free software. In fact you couldn't even get them to listen to free software or anything about it. They'd hear it, they think share ware, they would be extremely confused, it was necessary to come up with a little marketing spin so that's what was done. So it's not like there was a split in the movement or anything like that, it was just a marketing move. Why is open-source the way to go here? One of my favorite reasons is that it cannot be killed. Those of you who write code, if you ever written code for a company, you know that the company owns the code and if they want to use it for something, they use if for something, and if they don't they lock it up in a closet and it disappears forever. You can't do anything with it, nobody can do anything with it and given the way the IP laws work you are not even supposed to use the ideas, those ideas may belong to your employer now. It's very hard to do those kinds of divisions of your brain, so that oh I did that at IBM in 1989 and I did this at HP in 1990, that's not how the human brain works. This doesn't work very well in reality and when you think of the amount of intellectual property that's locked up in closets now and I love Xerox Park, but let's take them as an example. How much software developed by Xerox Park is just sort of sitting around doing nothing? The answer is, I think, is a huge amount and for no reason. What I think would make sense would be if in fact a company, any company, has software that they're not going to anything with. The reasonable thing to do, both for their own benefit and for the benefit of their employees is to open-source it. It helps the company's reputation. Another way it helps is by helping reduce the chance that some other company is going to patent the technology that this company has already come up with. If you have it in the closet nobody can tell that you had it earlier. The prior art is not out there. The way the patent system is going now anybody can patent anything, whether it is new or not, it happens all the time. So if you want to protect yourself and make sure your company will always have access to that technology, you're not using it in a commercial project, open-source it. Get it out there. Your programmers will love that. It will be a great recruiting tool if you can tell them; look if we are not going to commercialize it will be open-sourced. This helps the programmer's reputation and that's what motivates the very best programmers. Money? Yes they can get lots of money, but these people are artists. They love their work; they want to share their work with other people across company lines. How can
you enable that? At the very least you can tell them look if we don't use it we'll can open-source it. How do you motivate people outside your company to help you with your open-source project? There is the reputation value. Let's say you've got a company Red Hat, lets take Red Hat, and how do you motivate people outside Red Hat to help Red Hat with their software? Well you may recall when Red Hat IPO'd they used the friends and family stock offer to reward key people who have helped them. So in a four-profit environment it is possible to give, not just reputation value, which the programmers value a lot, but actually cash, which is always nice, and is a symbolic of acknowledging their contribution, which they also like. Why am I focusing on the commercial side? The reason is that you can do a great open-source project, but unless it's eventually commercialized you don't have the marketing dollars to get the attention of people out in the world to get them to use this thing. You can have fantastic software but think of the volume of information that comes at you everyday about new web things. It's totally overwhelming; you have to fight it off. So how can you ever find out about good open-source software? Somebody has to have some dollars to get the news out to you. Another advanced of doing things in an open-source way that I don't hear mentioned very often and I think it's because companies don't really like to play this up, is that by making your product open-source you can pull help in from other companies, regardless if whether the other companies want to or not. The license of the code is set, it cannot be changed, it can't be stolen, so whether your employees can work on the code if they want to, which they are going to do whether you like it or not and the project is benefited no matter where these employees go and given the way they circulate so quickly now you can't do a project where every two years or six months you loose all the information in a human beings brain, you've got to have access to that. So for the project, if not for the employers, it's a big pay off to have continuing access to that expertise. I mentioned that open-source projects, I believe, get the best architects and programmers. I think there is no question about that, because they are getting reputation value of payments you'd have to pay them a lot more in cash to get them to give you their intellectual property, especially with the knowledge that you may lock it in a closet and their reputation may never benefit from this. A lot of you have probably, I know some of you here in the audience, have written books, if you've read a book contract the way it reads is if the book goes out of print the rights revert to the author. Right? Doesn't that make sense? If the company who owns this thing is no longer interested in it do they get to just lock it in a closet? No, it reverts to the author and you can see why that would make some sense with software as well. That's in essence what open-source ensures that the creator of this work will never loose access to at least the reputation value, he or she may loose everything else but the reputation value will stay. Open-source seems to be more likely to follow existing standards out in the world. You remember when there was the big Netscape Microsoft battle on the browser technology? The goal there was not to support existing standards; it was to do great new features regardless of what the standards were. Part of the reason for that is the standard organization was to slow and I think that's why the whole model of how do you establish standards and software is changing. I think instead of having a bunch of people sitting around a table or online debating making theoretical academic arguments, instead it will be an open-source process where the way you would establish a standard is by writing great code and showing, saying alright this is good and proving that way. If only all standards could be done that way, it
would be great. I think for that reason open-source efforts tend to be much more likely to become standards. They're in conformance with existing standards and they move faster. I think open-source projects move faster, they have access to many more minds; the openness is critical to having a lot of input from people whose employers may not even approve of this. 

One common advance of open-source is said to be fewer bugs, that's kind of an obvious thing. If you think about a large software program, think about for an example Windows 2000, I've heard that it has fifty thousand lines of code.

AUDIENCE: Fifty million.

PETERSON: Fifty million, thank you, see I knew I would need your guys help sometime. When that number was announced at the first open-source developers day there were all programmers in the audience and someone said fifty millions lines of code and they went "ah", there's a gasp, they were appalled by this number just the sheer magnitude of that complexity. How can you get that to work? We will see, maybe it will work great, I don't know. The question, you think about that much code and you say how many people have actually looked at each piece of that code? How many? The one person who wrote it, okay, that's one. Anybody else? I don't know. At least with open-source there's a chance that the code has been reviewed by multiple people. It makes it possible the thing is going to work. There's a question of speed of development. It's important to realize you have a greater speed of development for a given level of quality. Now there has been fussing about the speed of development in Linux. Well, that's because the technologists are in charge, the marketing people are not in charge and so the technologists are going to wait until their ready and until then it's just going to wait. There is no one saying we have to release this on March 1 because there is this trade show, it's not how it works. There is no one saying well we're going to make up ten new dumb features to put on the outside of the box so we have something to write in a press release so that somebody will upgrade to this new version, it just doesn't happen. 

For our open hyperdocument system perhaps one of the most important advantages is that in theory at least we can attract diverse funding sources to the same project across company boundaries, Doug likes to talk about government, I don't like to talk about government, I would never go to government for this but okay we can disagree on that. One thing I think we can agree on is that, how many of you have heard about this Red Hat center for open-source? Not so many, there is a reason for that which is it was announced in November they say that they're going to have eight million dollars to promote the open-source concept, even beyond the software development borders. In other words you can take the open-source theory and say well what if you applied it to politics? Business? The open management kind of thing, it's kind of like open-source in business. That's eight million dollars that I think that the hyperdocument system project might want to ask for. They haven't been doing a lot of advertising of this yet, I don't think it is really set up, but Mark Ewing, anybody here know Mark Ewing of Red Hat personally? Okay, he's the man, we have to get to Mark Ewing, and he's going to be running it. Now in terms of selection of the open-source license, assuming we do decide to go open-source with this, it's not my area of expertise it's a complicated issue. The actual licenses themselves are pretty straightforward but the question is what kind of social results come out of it? People will argue that the ganu public license leads to less forking of the code, that's a possibility. The Berkeley license, which we use for our crit project, allows more flexibility in what happens with the stuff after you release it. Is that a good thing or bad thing? We'll be debating that; it's not really my area. About spreading the open-source concept to other areas I think we have to say we are doing that. This project that Doug is proposing is not just a software project, it's beyond that. We want to take these concepts and push them. In addition to the Red Hat Center for open-source, another organization that tries to do that is Foresight Institute, my organization. We will be having a meeting May 19 through 21 where we will do some of this; it will be in Palo Alto. If you want to go to that meeting or interested in details, don't wait because last year this meeting sold out about three months early and it was very difficult to get in. So if you are interested come and give your e-mail address either to me or to Tanya and we'll get you that invitation. 

ENGELBART: What is it again for? Explain again what the meeting is about. 

PETERSON: Doug wants me to explain what the meeting is about. That's kind of hard. Foresight looks at coming technologies, powerful technologies such as nano technology and tries to look at ways to maximize the benefits and reduce the risks associated with them. I think that there was a talk by Neil Jacobstein earlier in the series that talked about some of the down sides and how to head them off in nano technology. That's the kind of thing we talk about in the open hyperdocument system is one of the tools that we are attempting to do. We did a very tiny piece called crit; I think Tanya Jones talked about that earlier in this series. So we are working on a lot of the same problems that Doug is. So that is May 19 through 21, and if your interested don't wait, please come see me or Tanya. That's it.

ENGELBART: Do you want to stay for a few questions? 

PETERSON: Yes, are there any questions? Comments? Corrections? Yes, Shane.

CHANG: Open-source tends to lead to a very mixed issue; I'd like to hear your views on liability? 

PETERSON: The question is about liability and liability issues with open-source software. Well it's an interesting question and there's a couple of ways to come at it. One is you can say, all right whom are you going to sue? In the case let's say Linux, you can sue if you don't like Linux, if there is a problem with Linux for you, doesn't work, it blows up your plant, whatever. If you got your software from Red Hat you can sue Red Hat and I think that because it came out from Red Hat I don't think the question of who wrote that individual line of code that caused the disaster would come up. I think it's Red Hat that would be legally liable for that, which means Red Hat better feel pretty comfortable and confident in the code. A more serious problem, I think, is patent liability in software and that has not been resolved. I think the answer there is not to let it inhibit us on open-source, the answer is to try to reform the software patent system itself and actually that's one of Foresight's charters, one of our current missions is to take that on. I know it's very difficult, were up against some very large companies. On the other hand we have some pretty good tools we can use on this. It's a very ambitious project to reform the patent system for software but we're going to take it on. The last point on liability is that we at Foresight have taken the attitude of well we are not going to let fears of liability inhibit us from doing what we know is right, what ever it is. The thing is these days there is so much legal liability for everything, it's like you take a step and you can be sued for something, who knows. You can either let that intimidate you into doing nothing, or you can go on your way and do it anyway. Take Ebay for example, Ebay actually has always been violating California laws about auctions. They were from the beginning. Did they care? Did they even know? I don't know. Now they've got the money, now that they succeeded, they've got the money to deal with all of that stuff but if they have worried about it at the beginning they might have never gotten to be where they are. The whole liability thing is such a mess that I just refuse to let it paralyze me. 

AUDIENCE: When you started discussing open-source it struck me that the ultimate open-source for humanity is perhaps the human gene project and that reforming the patent system for software could very well be spring boarded from reforming the patent system over treating of human genes. I find it personally offensive, I think if a company could own rights to a gene in my body where they didn't even know it existed when I was born. So there is something morally offensive to me about the idea of patenting human or any other genetic material not generated in the laboratory. It would seem to me that there's a lot of leverage in pursuing that route to challenge software patents as well. 

PETERSON: Yes, I agree that's on our list too, were going to go after that one as well because I think it's just ridiculous that they are permitting patents of natural genes found in nature. I mean new genes, okay, we can talk about it, natural genes, I'm sorry you didn't invent them. 

AUDIENCE: You can talk to the Pope about prior art. 

PETERSON: Back.

AUDIENCE: I understand that in at least in California that unincorporated associations every member in the association is legally liable. I'm not an attorney, but I would prefer to see anything that I joined to be say a non-profit corporation just for that reason. 

PETERSON: Yes, in terms of actually one reason that I can afford to be so cavalier about legal liability is that Foresight is a non-profit corporation and we have excellent directors and officers insurance that covers the volunteers as well. So if any of you want to do anything that's a little nervous making come see me and I'll refer you to my insurance agent and you can be covered for this too. Actually the non-profit corporation form is very useful for dealing with liability. Yes, sir.

AUDIENCE: Can you talk more about the difference between the Sun Microsystems's model and the open-source model? 

PETERSON: Okay, yes. The main difference between the Sun Microsystems's model, called the Sun community license and the open-source model is that Sun wants, and I may have this wrong and stop me if I have it wrong, but I believe Sun wants final say on code that is being redistributed. In other words, if you want to take Java and go off in your dorm room and change Java, you can do whatever you want to Java. If you want to take Java and start redistributing it commercially, number one I think Sun wants to see that code in advance and approve it. There concern is compatibility issues. You can see why they have compatibility issues concerns; they just won a lawsuit against Microsoft for exactly this issue on Java. So they have issues here. The other thing is I think if you do it commercially and make money you have to give them a royalty. So the main difference is that under the open-source model you really do relinquish control to the process, you trust the process. Under the Sun community license model you relinquish control to Sun, and you trust Sun. Now there are a lot of people in the world who feel more comfortable trusting Sun then they do the open-source process. For example, I think Bill Joy believes that big companies feel more comfortable interacting with another big company then they do, as they might see it, abandoning themselves to this unknown not well-understood process. I can understand that. He might be right. I think for our purposes for open hyperdocument system that's not an issue, but for Ford who knows what their incentives are. 

AUDIENCE: I've actually been working on Sun in terms of with Genie, which is another technology, which actually pioneered the scuzzle Sun community source license and basically what you are saying is right. But the interoperability is a real key thing there that the idea of the technology for Java, that your Java pro won't run everywhere and some companies like Microsoft would do their extended embrace and change it so you can't use it, you would have to use theirs on their platform. So that is a major concern and I think it would be a concern also for an open hyperdocument system in that if you want to have multiple organizations being able to have their documents read by each other, they need to be interoperable. The other difference I would say between the two is one that you actually brought up between yourself and Doug, which is the governance model; the classic model for an open-source project is hopefully benevolent dictator. You just trust that the person who held the original vision for the software is going to be the one making all the decisions and you just sort of trust that's going to work out and if it doesn't then there's a forking in the code and sort of a revolution happens. The Sun model right now has been one of a community governance system where the community members do have a say. So it's not Sun controlling it, it's Sun saying we want to make sure we've got a process here where change can happen, even change Sun doesn't particularly care for, but change will happen inside the governance of the community so the whole thing will move as a hole. 

PETERSON: That's a very real concern. The interoperability thing is a very real concern as for whether we have to go to the Sun community license model. I think the reason people think of that as sort of the benevolent dictator model, one reason is because of Linux, and Linus Trovals and that model. I think, and again help me out here, but I think the Apache model is different. 

ANDREW PAM: Among others, yes. The Apache model is a group of people who form a kind of, collective if you will, and there are a number of other significant open-source projects that do not have a novelty title ship model. So even though that is certainly quite common, it's by no means the only open-source model. 

PETERSON: I should mention, I'm going to wrap it up. Doug says I'm going to wrap it up.

AUDIENCE: Benevolent dictator.

ANDREW PAM: Pearl is another classy example as well.

PETERSON: I should mention that it's possible that the Sun community license model given that there're trying to develop the same kind of thing where you have a real community that is not totally Sun dominated, it's possible that as time goes by we may find that the way that it operates isn't that different from the way the Apache model or the Pearl model operates. That will be good if it happens and we all hope so and I certainly think, as I said before, I think Bill Joy and Sun deserve a chance to try out their model. 

[  principal lecture]

 ---