>> Kamal Jain: It's my pleasure to introduce Mike Schwarz he's going to be at Yahoo Research. I think he had been to many of [inaudible] institutions also. Besides Yahoo, he had been a professor at the Harvard University, then he went to Stanford and then Berkeley and then finally to Yahoo. And today he'll present us about something about open source. >> Michael Schwarz: Thank you. So let me start with a bunch of disclaimers. So disclaimer number one, the title of this paper is misleading. It probably should have been called Open Source, a Solution to Hold-Up Problems, Half a Century of Public Software Institutions and not the other way around. And this is the way we'll present this paper, so start with the second line and then I will continue with the first line and start talking about the history that actually illustrates how open source solves hold-up problem. But first I will have to explain about the hold-up problem. So rationally I anticipate that even though I serve the second part, the history part is something like 70 percent of the paper and it's really interesting history. I real recommend reading it, because it's something like 15 pages and it's a hot of fun. Nevertheless, I anticipate that I will probably spend 80 percent of my time on this part and I will probably end up completely running out of time and spending about three minutes on this part. But hopefully the first part will be sufficiently treating that they will actually want to read about the history of open source. So the other disclaimer is that I am normally not a nervous presenter, but this time I am because I'm talking about something I know virtually nothing about to a group of people who know about this something tremendously more than I do. So I'm an economist, I'm an economic theorist, I prove theorems about auctions, about matching, about market design and I also design markets where hundreds of millions and billions of dollars actually transact. So that's my day job studying markets and actually designing them on practice. My job has nothing to do with software, and I know very little about it. So this paper is very much kind of whole bit, kind of an exercise out of curiosity and here I'm talking about something I don't know much about. So the disclaimers. As an author this paper is very unusual paper for me, it has no formulas, no theorems, none of that nature. So I'm just not sure -- so finally another important disclaimer, so obviously you probably know that Yahoo uses open source quite heavily and also contributes to open source quite heavily. And none of the views expressed in this paper have absolutely anything to do with the views of my employer and also I don't have any inside scoop on the Yahoo strategy vis-a-vis open source, so all the opinions are entirely my own and those of my coauthor, Yuri Patev [phonetic], who is a PhD student in the school for information at Berkeley. He is primarily a sociologist. He is also an occasional open source programmer. So with all those disclaimers, let me move on to first slide, completely irrelevant slide for this audience where I'm trying to explain what open source is. I don't have to explain this to this audience I believe. I suspect that everybody knows that there are two basic types of open source licenses, permissive and copyleft. Does the everybody know the difference? No? Who knows the difference between permissive licenses and copyleft licenses? I should talk about it, because it's fascinating. So generally speaking open source software is understood as a software where when you get the software you also get the source code. Together with the source code you get a license that allows you to modify those source code. That's what open source is. But then there's a second issue. And you're also allowed to share whatever you did. You modified it, you created the new program, perhaps you take this one. You can share it. That's also part of being open source. You can modify it, you can share. But then there's a question when you do share this software -- by the way, sharing didn't mean you share it for free. You're perfectly in your right to take an open source software, maybe modify it a little bit, and then sell it to someone. That's allowed. That's fine. That's allowed by any type of open source license. So the restriction that comes, however there's a licensing issue with permissive license you can do whatever you want. You can take the software product and then you can modify it, you can sell it to someone else and you can sell it to someone else with whatever license you please. That's why it's permissive. Do as you wish. Do as you please. The other type of license is copyleft. Copyleft is quite different and quite different, I have to say. Again, you're allowed to modify and share and sharing doesn't mean sharing for free, you can sell it to someone. But as you share or sell that software, you have to give it to the new person, to the person who whom you are selling it, with exactly the same license with which you received it. You couldn't -- you are not allowed to change the license. And that means that that person who will get your software, that software from you, that that person would have a legal right to sell that software or to give it away to free to everybody in the world. So as you can see, there's a tremendous difference because basically you can take an open source product with a permissive license, build the commercial product based on that and profit from that commercial product. If you take software with a copyleft license, you can build a product based on that, but once you start selling that product, you have to allow your users to give away that product for free or to sell that product to everybody in the universe, which basically makes it impossible to make -- to sell it for a high price, right, so -because if your price is at all high, people would start undercutting you or even giving it away for free to everybody in the world. >>: [inaudible] I have a hypothetical question. [inaudible]. Let's say somebody takes the open source code and put -- I mean they hide things in the code which are advertising based, then would it [inaudible]. >> Michael Schwarz: That's an excellent question. And selling is not important here. Share -- you are still allowed to sell it for money. It's fine. So it would be considered sharing. And when you share, you have to preserve the license. That's all. So you are more than welcome to take an open source product, to put a bunch of advertising into it and to start giving it away to people or selling it, as you please. But you have to preserve the license. So once I got it, you have to give it to me and I have to have a right to A, modify it, and, B, give it away to the whole world for free. >>: [inaudible] I mean makes [inaudible] so that it's [inaudible] the same but it looks different. >> Michael Schwarz: That's always true with copyleft license. You can even sell it on those conditions. That's fine. But you have to allow people to distribute and modify it. And ->>: [inaudible]. >> Michael Schwarz: Huh? >>: [inaudible]. >> Michael Schwarz: Not with permissive ones, with copyleft ones. So by the way, if you will wait, assuming we will get to the history part, the copyleft license was a truly brilliant invention that played a very important role in development of open source software. So let me give you a little bit of summary of how incredibly important open source software is today and sort of given this audience I probably don't have to kind of [inaudible] open source is very important, on the other hand, given that I am in the Microsoft right now, maybe I do have to convince you that open source is in fact very commercially important today. So the staggering quantity of open source software. So for example, one can do absolutely fine relying exclusively on open source products, from operating system to browser, to office, to statistical software, to development tools; and in fact, my coauthor happen to be one of those people who over the last three years used only one known open source program, and that happened to be Adobe Flash that is free. Everything else that he used is open source, and he doesn't seem deprived. So that's kind of interesting because ->>: [inaudible] getting around in the wheelchair and do everything. >> Michael Schwarz: That is true, that is true. But there's a difference because he and many other people like him don't feel particularly deprived, right? So for example if you were to limit yourself to say open source music while there is not that much around, not that much of open source music around us and what is there is very amateurish, right? So all this sort of best stuff is not really open source. So in fact, what's interesting is that free software is high quality. So oftentimes we think of free stuff as kind of this is stuff for poor people, right? So people who couldn't afford -- so people who couldn't afford the real deal, they go with free alternatives. And that might be the case in many instances. Open source is not it. So in fact, open source software is quite high quality. The competition of open source and proprietary software is larger and not on price. This is what's quite interesting. When I starting writing this paper, I thought that this was about price, about money. But it's not at all the case. So, for example, many users of Linux not only made the decision -- are not only quite often and could easily afford Windows, in fact they paid for Windows and they deleted it and installed Linux operating system. Also Mozilla Firefox and Internet Explorer are completely with each other. Mozilla is open source, Internet Explorer is not. And both of them are free, and the other Mozilla is doing very well. So again, the this is clearly not an instance of price competition. So in fact, also an interesting observation is that the most use of open source by individuals is in fact the tip of the iceberg compared to use of open source by corporations. As you will see two bullet points later, it is really the corporation's, the very wealthy corporations and not the individual users who are relying most heavily on open source and that again suggests that it is really not about price, it is about something else. So for example, let me give you just couple of examples that illustrate how incredibly important open source is. The world's largest supercomputer, IBM Roadrunner, and the world's largest database, our own, are based on open source software. >>: [inaudible]. So what software on IBM Roadrunner is open source? >> Michael Schwarz: That is an excellent question. And to that question I have to confess. I don't know the answer. I have a source that we quote in the paper, and my coauthor, who is much more programmer than I am, could give you a much more precise answer to that. So if you want to know the precise answer, drop me in the mail, and I will get you a precise answer. But I believe it goes beyond the operating system. I think sort of the whole sort of -- so I presume this -- I'm guessing that Roadrunner is probably not a single CPU but huge, huge, huge collection of boxes. So I would be guessing that basically the programs that allows those boxes to work together than that part must be open source. That would be my guess. But I couldn't name the -- and I premium the operating system would also probably be Linux. But I could -- given that it's IBM, I'm sure the operating system would be Linux. >>: [inaudible]. >> Michael Schwarz: Huh? >>: If you modify it and you don't distribute it, you still call it open source? >> Michael Schwarz: So that's a very good question. No. If you are -- so the different types of open source licenses are but generally speaking with sort of Internet created sort of new type of usage of open source software because now a lot of software is used by creating value for the user without actually giving the software to the user, by providing a service. And that's the business that Yahoo is in, the business is Facebook in, My Space, Google, we are all in the business where they are not software companies in the sense that we are not selling any software to the user, user doesn't see any of our software. But we are providing services. >>: [inaudible] or these companies found a way to circumvent the copyleft license by kind of not really giving the software but in a sense getting around this. >> Michael Schwarz: Well, I -- so actually I don't think that those companies circumvented copyleft license. I think that was actually original intent of copyleft license. Because the original intent of open source licenses was that you take open source software and then you use it for a business, for purposes for doing whatever you want to accomplish, and in the process you modify it in a way that are useful for you for doing the job that you want to get done, and then you're not required to make it available to others. So this is not a part of open source license to make it available for us. And that was the intent. That was not -- that's a feature, not a bug. However, you may choose to make it available to others and in fact if you would look at sort of Yahoo and Google and My Space strategy in this space, some My Space is still a baby in this space, but if you look at Yahoo and Google, Yahoo is -- and Google and many other Internet -- and IBM for sure are very heavy users of open source, and a lot of things that are done by those companies with open source stayed with those companies. [inaudible] you don't share it with the public. But a lot of this is shared. So if you look at the least of contributors to open source you would also see IBM and Google and Yahoo and Red Hat. It didn't mean that everything is shared, but a lot is. And some isn't. And I don't know the quantities. Right? And that's how it was intended -- intended, I believe. >>: [inaudible] where one could be featured that [inaudible] I took an open source, modified it and created the binary. One situation is I put binary on let's say [inaudible] computers for him to use and other way to use binaries, I use [inaudible] use my own computer remotely to use the same binary. Why one would be allowed and one would be a feature? Why would one be allowed and other not? >>: So the point is I think we disagree with [inaudible] inventors of copyleft didn't envision the scenario where people could use the software without having it, and so I think it is a bug. >> Michael Schwarz: So I think ->>: [inaudible]. [brief talking over]. >> Michael Schwarz: So I think to some extent this is a bit of a philosophical question. But there is an answer to this question, and the answer is a new copyleft license. The new license that basically says well, not only you have to share the source code, in case you share the software, even if you provide the service over the net using that software, you still have to share the code. That license have been developed, I believe. And it's very unpopular. It is not being widely used. And part of the reason is obvious. Google, for example, and Yahoo are very active contributors to open source. And as far as I know, none of the Internet companies wish to either use or contribute to software on those terms, and yet they're perfectly willing to see -- to both use and contribute open source software with a different license. So that simply suggests that in some very fundamental level given that we see sort of open source is really flourishing on those terms but not with other license. That seems to suggest that -- I'm not saying that that other license is somehow wrong or evil, I'm just saying that both have room in the universe and at least for now the current one is succeeding quite nicely. >>: [inaudible] so that's why [inaudible] like I don't understand what meaning of given by [inaudible] if I take a binary and burn it on a hardware and sell that hardware [inaudible]. >> Michael Schwarz: This is an excellent question. And I have an answer for that. The answer is -- and in fact, I -- so the short answer is you could not do it. If you take a binary and burn it on a hardware, together with that piece of hardware, you've got to give the source. How do I know that? Well, I know that because one of the most common Internet phones called Linksys, that's I think one pretty much one of the dominant players in this space, and of course there's ones the way of the future, I have one of those in my office and ten years from now or even five years from now probably every corporate office will only have Internet phones, right? So this is very kind of everyday technology. So that company build that Internet phone, and I think it's dominant player in this space. They build that form using Linux operating system. And the other operating system happened to be -- happened to be copyleft, Linux's copyleft. So then they started distributing those phones, very prosperous company as far as I can tell, and they were brought to court. They said look, you are distributing the software, you're distributing the binary, but without the source code. You are if violation of Linux license. And the court said, well, that is correct. And the argument was software is often distributed together with hardware. For example, I -- so you see, now I know that I should be right now at the daily marketplace sync up, but I am not. So, yeah, Outlook respects my privacy. So the reasoning of the court was, well, let's say I take a piece of hardware floppy drive and I put a binary in there, I give you that software and the floppy surely, it's very well established precedent that if I give you a software on the floppy in a binary form now you have to also give me a source code upon request. So. >>: [inaudible] but if I find it on the phone and I don't make it easy to read binary then I'm not actually giving you binary. >> Michael Schwarz: That was the other side was arguing. They lost. Perfectly good argument. But it's subtle. They lost. So learning from this lesson when Apple decided to develop the iPhone, they also didn't go out writing an operating system from scratch. But they went with standard distribution UNIX, which happened to have permissive license. Because I believe it was developed before copyleft was invented I'm guessing. So -- well copyleft was also in that. But I believe that that was the timing. So iPhone has a proprietary operating system that is based on open source software with permissive license. On the other hand, Android, the Google phone, is action open source operating system based on Linux. In fact, I think as far as I can tell received wisdom that Linux is superior to standard distribution UNIX which is by now a bit outdated. And I would be guessing that the only reason why Apple opted for that is they want to keep the operating system proprietary. But I'm not sure. Maybe there were other considered technical considerations. So this slide is taking tremendously longer than I expected, by the way. So ->>: [inaudible]. >> Michael Schwarz: I know. The time slows down, it speeds up, depends how you look at it. So a couple of other kind of data points. Most computer languages are open source. And that includes some of the most popular languages. So Java and Python and -- Java is probably the most popular computer language in the world the compilers are open source. So interestingly enough, they started very differently. I think Java was originally from proprietary language, it was not open source. And then Sun opened it up, completely opened it up. And they were gradually opening it up more and more and more and then they opened it up, completely opened it up. And they opened it up while they were still the most popular language in the world. And I think they opened it up because they were sort of feel the pressure, they were afraid that python or some or language would come from behind and overtake that -- overtake their standard, and they thought that sort of making it open source would make it more appealing. So it was a defensive move as far as I can tell. I'm guessing, right. God knows. And that's interesting because the story played itself Fortran and COBOL. That's how IBM was opening Fortran, back when in the '50s. Maybe early '60s. So the history has a tendency to repeat itself. So it's interesting that if you look at compilers, computer larges, essentially almost all of them are open source. There are some exceptions. For example, C has -- it's actually Microsoft has a I think C -- is it Microsoft? I'm not sure. But C has some proprietary compilers. But those have plenty of open source initiatives. But essentially almost everything like nine out of ten top languages are probably open source. You will get computer games. Nothing is open source. All the popular computer games are proprietary. So in fact that's not true. There are thousands of open source computer games. It's just none of them are successful. So this is quite striking. Pretty much all the successful programming language, with very few exception, right, are open source. And none of the games are. So there must be some reason why it is this way. And what this paper is about, what this paper asks is first of all why is open source so successful? Because as an economist, I can see all sorts of reasons why it isn't be successful, so I want to see some reasons why it should be, and second, why does successful in some places but not in others? So a couple of other observations about open source. Open source completely dominates web serving. And I know Microsoft is a player in web serving. But of course Microsoft have much smaller market spare than Apache servers. So all the big Internet companies are using open source solution for web serving. But Microsoft has a still considerable market share there. But certainly much smaller than open source overall. And also, infrastructure of Internet companies, Amazon, Facebook, Google, Yahoo, many, many other smaller ones is incredibly heavily based on open source, right? So we see that in certain areas open source completely dominates and in certain areas it's almost nonexistent. Yes? >>: [inaudible] open source. But Google makes, you know, 95 percent of their money from search, right? >> Michael Schwarz: Right. >>: Their search engine is not open source, right? I mean, I can't get their proprietary code ->> Michael Schwarz: Well, of course not. >>: The way they do page ranking. >> Michael Schwarz: Of course not. >>: So it's mean, it's nice of them to, you know, have all this open source source code except what makes 95 percent of their money. >> Michael Schwarz: So first off, so first off I'm not making ->>: Say Google is 5 percent open source and 95 percent closed source. >> Michael Schwarz: So first off, I'm not making any value judgments here. I'm not saying that, you know, Google and Yahoo are nice and some other companies aren't, but second -- [brief talking over] >>: [inaudible] is happily ->> Michael Schwarz: No, but, but, second of all, here's the [inaudible]. >>: [inaudible] runs on [inaudible] it didn't let everyone see what their ranking algorithm precisely is, right. >> Michael Schwarz: Well, wait. >>: So it's open source. >> Michael Schwarz: So you see, you are -[brief talking over] >> Michael Schwarz: You are thinking of them as a competitor? >>: No, I'm not thinking of them as a competitor. I'm thinking that to claim the Google -- to claim that open source is successful because Google uses it I think is wrong, because where Google makes their money, which is all in search, that's not open source. So let's look at what the company -- where the company makes its money and then say that's code that pushes that -- the money maker, right, in open source. >> Michael Schwarz: Well, so the key word here, the key word here is infrastructure, right? So I'm not saying that what they create -- that they're sharing everything with the world. Of course not. What I'm saying is that sort of the key enabling input for their technology is open source. I'm not saying -- I'm not saying that the ->>: I disagree. >> Michael Schwarz: That it does you any good. >>: No, it's not -- I don't think -- you know, I think, you know, that their key technology is their page ranking algorithm, and that's not open source. >> Michael Schwarz: So I'm thinking at a infrastructure level, right? So the operating system that ->>: [inaudible]. >> Michael Schwarz: No, no, of course they don't sell it. That's what I'm saying. As users. So you can think of Google in two ways, right? You can think of Google as a company that provides service, that creates value. And so let's forget Google. Let's say we are thinking about Chevron oil -- Shell Oil, right? So Shell is not in the survey business, they're in the oil business, right? But they also happen to be computer user. So if you find out that most of the survey that they use is open source you would say, oh, Shell happened to be a heavy user of open source, right? That's not where they make their money. They make their money in oil. But you would agree that if they ran operating system on the computers, they are writing the databases based on operating system search and you say Shell clearly is a heavy user of open source. They're also an oil company. >>: [inaudible]. >> Michael Schwarz: So all I'm saying here, all I'm saying here, is that all those guys are very heavy users of open source and that open source is critical for their success. I'm not saying that they're sharing any of the benefits with ->>: Open source is critical to their success ->>: But that's really ->>: Google could certainly afford to buy Windows licenses and it wouldn't even affect their bottom line. >> Michael Schwarz: That's exactly the key point. >>: Okay. >> Michael Schwarz: So why don't -- so that's exactly the key point. So not all those companies are poor. In fact, they are rich are. And usually when you are sort of in a business that's winner take all, you take the inputs, you know, the cost kind of, you take the inputs that would make it the best possible product. Right? That's what you want to be. You want to win in the marketplace. And then you just need to share it with the supplier about the right price, right? So if someone has something that works better for you, you will just figure out the deal hopefully, if possible, and you would use that. So it's striking and interesting that does all those companies are opting not to go with commercial product, they're opting to go with open source. If you look at other things that they buy, you know, they're not that cheap. You know, they buy Aeron chairs. So why is it that they choose to use open source inputs there? Right? So that is the key point here. Why open source -- why do they choose to be consumers of open source? Right? >>: Started out as ->>: Let him ->>: Okay. That's the point. I mean why does Google use, you know, Linux, they have to sort of maintain themselves as opposed to buying Windows which we would maintain for them. >>: The question is when did they become so successful and rich and whether it's at that point did it make sense for them to switch their entire infrastructure -- >> Michael Schwarz: [inaudible] it's not a again, because when they are little, then there's even more reason if you are little, you are trying to succeed. So, you know, Google doesn't build their own chairs. And when they were little, when they were a poor startup, they just went out bought a bunch of Aeron chairs for 600 bucks apiece? Why? Because when you try to make great software you need good chairs for the butts of your programmers, and Aeron does the job and it's fine. So you would think that you know certainly in the interest of Microsoft to get startups hooked on Microsoft technology. So surely there ought to be a possible deal between a startup and the proprietary vendor ware, you know, you would use our software, you know, for the first -- until you start making money you don't have to pay us or something like that. You can figure it out. But that's not the route they went. So why is that? Nobody is going that route in the Internet startup field. So why is that? So that's what this paper is about. So let me sort of flesh out what are the questions that I'm going to try to address here. So I'm -- economists thought and wrote a lot about open source. So that's including Microsoft chief economist [inaudible] open source and most of the issues that economists are concerned about, what are the incentives to contribute to source? And I wouldn't talk much about this. This is not the question that we are going to address. And in fact to some -- so this is definitely not a question why people contribute to open source. There are lots of good reasons, including just for fun. That's how founder of Linux called his memoir. And so by the way, if just for fun were the real answer, then you would expect there would be lots of successful open source computer games and there are lots of open source computer games not of features successful. So the real questions with we are trying to address is why is open source dominate some important sectors of the software and not others? So that's what we are about here. And the -- so it's not that surprising that people contribute to open source, right? People do engage in a lot of activities for fun, right? People create music for fun. But how much, you know, you don't see Paramount Pictures taking some amateur song that's available copyright free and making it into sound track of the movie, right? So all that -- but yet you see software companies doing the Internet companies doing exactly that with the software. So why is it happening, right? So bonus questions of Stallman, who is I think Richard Stallman is basically the founder of open source movement, he talked to his -- so this is a bonus question. Why did Stallman -- what did Stallman meant when he asked his supporters to think of free as free speech and not free beer? When thinking about free software? So assuming that I will not run out of time, and I'm sure I will, I will give the answer to this question. Otherwise, you'll have to read the paper. So now let me sort of proceed -- so there is no really model in the paper in the sense that we don't have formulas or anything of this kind but let me sort of give you the fundamental assumption that kind of drives a lot of things that are happening in this paper. So one observation which is kind of obvious is that one fundamental difference -- so one obvious difference between open source and the proprietary software is the price. But there's one more difference. Open source software is very highly modifiable about it definition. You can take it, you can modify it, if you know how. Binary software is impossible to modify. So if you take the same program in binary or open source form, in some sense open source without regard to price is a superior technology. It's superior in the sense that it gives user, certain users, sophisticated user lots of options that proprietary software doesn't give that user, and that gives us a key why it so happen that it is the most sophisticated and wealthiest users that opt for open source. Those are the guys that really value that flexibility. But that's going to be only the beginning of their story, not the whole story. So you can see that sort of intrinsic superiority of having the source. But it didn't have to be open if you think about it, right? So there's no reason why sharing the source with the user should mean making that source open in the sense of open source swear. I can share my source with you but give you a license where you're not allowed to share it with other people, you have to pay me additional royalties if you make modifications and blah, blah, blah. Right? So we could man those kinds of places in agreements. They're very rare. It has been tried. There are some precedents. There are some companies that do that. But the reason that there's a very difficult is contractual costs become extraordinary high for a lot of reasons. It's just too easy to cheat, too hard to monitor, too hard to agree on things. So as a first order approximation, we can assume that given the user code, this is technological restriction that once you shared the code you have to make it free. Legally it's not quite the case. You can write a contract that says it's not, you're not allowed to do lots of things with it. But it's so expensive to enforce in court that the most cases is just not worth it. Also, it expose both sides to a risk. I may be very hesitant to touch your code if I have to sign a paper that I'm not allowed to do lots of things because later you can bring me to court accusing me of lots of things and charging me for all I -- and suing me for all I'm worth. And that happened in the past. So there are real concerns there. So I will assume, and that's not true, it's first order approximation, that giving a user code is isomorphic to just making an open source product. Which is not entirely true in practice. There are counter examples, but they are few and far between. So that's the key observation that will be driving my story. And I believe in my heart that if it were possible to write and enforce contracts along those lines where I can say, look, I'll give you this code, but we have a complete contract that the totally reasonable that would govern our behavior with regard to this code for all kinds of future contingencies and you know it's cheaply and easily be enforced then there wouldn't be open source software. Because then, you know, we will just agree on the reasonable price and I will get some very nice proprietary product from you and just pay you a reasonable royalty. And so I think fundamentally this is the technological restriction, the legal sort of lack of legal visibility of enforcement that created the open source movement. And I think this is a fundamental reason why Internet companies do not want to touch proprietary software. So I'll illustrate that -- so due to the fact that users can sort of modify software, it creates a lot of additional efficiencies, right, user doesn't have to communicate his needs to the vendor, right? So if I'm a sophisticated user, I want to change something. If I couldn't change it but only a vendor could, then there's this costly process of me communicating those needs to vendor, hoping the vendor would make a modification then presume the vendor wouldn't do it for free for me, I have to negotiate the price with the vendor, and that's the key variable. The vendor may potentially hold me up, right? Even though modification may be very easy, since I'm already so invested in this the product and, you know, let's say I'm Google. So it's very important for me to modify your product in a certain way and it's not that heart to modify it, but you know, let's say that product happened to come if a vendor who realizes that, hey, this client makes a lot of money, it desperately needs this modification and I can make this modification for a million dollars, I will charge you billion because the client is in such need of this modification. So that's a classic hold-up. A client has this flourishing business that critically depend on this piece and I could exploit this client later on and there's no way we can sign a contract before that that prevent this kind of exploitation. That's what economists called hold-up. So hold-up arises from lack of contract -- from invisibility of sort of full contracting for the future contingencies. So let me give you an example that illustrates -- and this is basically the key insight of this paper the way I see it. So let me give you sort of a description of what we mean here again. So let's say a buyer of software product A makes a complementary specific development by buying complementary software which is product B. That investment could be training or it could be creating other software, right? So we have an Internet startup. They take some software, based on the infrastructure but then around it they write lots of other things. They create this whole service. So that investment, the product that they created could be tremendously more valuable than the software that goes into the foundation of that company. So however, so basically the value of B, the investment complementary to A could be much, much more valuable than A itself. However, if -- it may be necessary to modify product A in order to reap benefit from product B, from the big investment. So for example you have an Internet startup, they started something out, it's going well, now suddenly they are growing and in order to scale their the product, they need to change something in product A. But they couldn't, because proprietary. Right? So that would be big problem for them. They would have to go back to the vendor and so on and so forth. And unless they have signed the contract with the vendor where vendor says oh, I'll modify everything at a reasonable price, how do you define reasonable price? It's impossible, right? The vendor would say, oh, well, you're such a successful company, just pay us an arm and a leg for those modifications. So that would put that Internet company in the position where they would be afraid to invest into other software that complementary with software A, the proprietary software. And that is what driving them to use open source. That's a reason for hold-up problem. So in other words, clearly with open source software there would be underinvestment into developing. Who wants to create open source, right? I mean, couldn't make money on it. So it is underdeveloped. But this Internet company may choose an open source solution, even if inferior, even if there's underinvestment into that product. Because if they choose proprietary software, they would be afraid to invest in the reasonable technology. And now they can go all out and invest in complementary technology and in the meantime keep building that product that sort of open source foundation of the product. They also may choose to share the modifications they make. And the reason that they would choose to share modifications that they make is that they may hope that other people would yet improve that bottom layer of the infrastructure and they would benefit from other people's modifications, so there may be direct incentive to share with some of the modification that you make to open source products that you use. So in other words because of hold-up buyers of proprietary software underinvesting in complementary products and if you make big investments in the complementary products it makes a lot of sense to rely on open source solutions. So let me give you my favorite example. And I love that example because it's not software, it's hardware. So there's a company called -- I think it's called [inaudible] software or hardware and basically this is an open source hardware product. It's a small company based in [inaudible], a startup I think the sales are about 50,000 chips. They make chips. And those chips are all specifications. Everything is open source. You can download it, you can buy those chips from anybody else, right, you can just take those specifications, send them to any other vendor and get those chips from someone else without paying a penny to this company. So this company makes money by selling those chips. But they make it possible for anybody else to compete with them. They have essentially no idea. When they develop an idea, they immediately make it open source. So why are they doing it? That's crazy. Well, they're clearly not crazy because pretty impressive for a startup from Italy to actually sell 50,000 units. What are they doing? So here is my theory of what they are doing. Who buys those chips? Why would you buy a chip? So they originally developed the chip because it was part of some -- part of some toy or an art -- some -- I'm actually not sure what they originally used, but it could be used for toys, it could be used for art projects, it could be used for gadgets. So it's quite a flexible chip. So think about it. Let's say you are a designer and you are trying to design something, maybe a toy, maybe something else, and you are hoping you kind of strike it rich, right? You are hoping to sell millions of units. You probably wouldn't. You probably make one unit then just flop. So you are investing a lot into designing this new thing. So if you are designing something around a proprietary chip, once you become successful and you start -- and you start putting orders for like 200,000 chips, the chip vendor first of all, may not even be able to provide those chips, well, we don't have capacity to build those, or they could say, we've got plenty of capacity, you know, we are Intel, but, you know, we can charge you an arm and a leg. So -- and if it's a small company, then you are particularly afraid that, you know, I'm the first big customer, I'm trying to buy 200,000 chips because I had built a blockbuster toy based on that chip. Well, they would immediately just take me for a ride, right, to sell 100,000 chips the price would be very high. So people would never go with a chip from a small company that is proprietary. Because they would be afraid that once they invested everything in product development and design they actually wouldn't be able to get those chips at a reasonable price. So as soon as those chips are open source this company pretty much guaranteed that they would never take advantage of people who developed technology based on the chips because the technology is open source. So that -- and that's I think pretty much the business model. So also there's an incentive to share innovations, because once I develop something based on this chip, I may want to modify this chip. I may choose to share those -- I may choose to share those modifications with the public. Why? Well, because maybe I'm hoping that the new version of this chip would be better yet and there would be not just my modifications but other modifications. If I'm say a toy maker, I'm not very concerned about competition. >>: [inaudible] but now when I [inaudible] it looks incomplete in the sense that consider the [inaudible] developing it could be a client site service, it could be some software [inaudible] that could be software site service for example search. And I mean, it's entirely possible for [inaudible] what if I need to make it more efficient. But the same thing would hold if you writing client type software on Windows, which you may want to treat Windows to become more efficient. So the dependency on the underlying layer is the same. I mean Google could pretty much could have run its entire search service on Windows. >> Michael Schwarz: So ->>: And it would be about the same. >>: [inaudible] would be the source. >>: No, no, but why do they need to do it because do people get source who writes application on windows in they don't get source. >>: How [inaudible] is it, how important is it? I mean you got a Citibank, right, you know, they're very happy with the performance that Windows gives them and they're interested -- and first of all, they don't have lots of OS and computer scientists in their building, so they would have to create a whole division. >>: Yeah, but that's -- >>: To modify Linux, right? >>: That's a subjective argument because maybe I'm [inaudible] and now it might be critical for me to access ->>: [inaudible] and you can afford the OS guy, right, if you need to hire an OS guy? >>: No, no, but I'm -[brief talking over]. >>: Modify Windows. It might be critical for me to modify Windows to make my browser the best browser ->>: Right. And his argument is if it is, if it's that critical for you ->>: Then I would like for Linux ->>: Then you would hire an OS guy and get Linux. >>: Yes, but people like browser for Windows. >>: Because performance isn't that important to them. >>: Yes, that's what I'm saying. Why in some cases performance is more important and other it's not? I mean ->> Michael Schwarz: Well, I think that's -- I think that's exactly the ->>: It's the nature of the -- I mean it's the nature of the application, right? [brief talking over] >>: Because they believe performance is important. >>: Usually just features. It's the features, right? It's the features, it's the interconnectivity. [brief talking over]. >>: And one of the things [inaudible] we are fast, we are fast, we are fast. They are not saying that their features ->>: Well, because they modified Linux. >>: Are they going to write the browser for Linux? They're writing for Windows. So it's the same argument, that the dependency, the mathematical -- >> Michael Schwarz: Wait a minute. So I'm not quite -- so remember -- so they did write the browser -- a browser for Windows, but so remember, I'm not claiming that the paper that I wrote explains everything under the Sun, right? So obviously there are Windows computers and there are also applications for Windows. And I'm not saying that it haven't exist or doesn't exist. Of course it does. >>: [inaudible]. >>: [inaudible] their infrastructure. I mean this is like a product. >> Michael Schwarz: Yeah. I'm not claim -[brief talking over]. >>: But you are making a difference between that infrastructure is -- in both cases it's not infrastructure which is important, but it is their ability to search searches. If you have some magical oracle, they would search searches, and they could have built infrastructure on Windows [inaudible] searches. Same thing as Adobe [inaudible] is to [inaudible]. And their [inaudible] service could both potentially. [brief talking over]. >>: There are some ISP vendors out there performs on Windows not accurate. >>: Yes. >>: So then they're stuck. So then they're stuck, right. They either put up with what they consider to be poor performance or they go out, they use Linux, and they modify the OS in order to provide the performance they need. >> Michael Schwarz: So also, I don't think -- so also I think the key issue here is not that it's just poor performance or good performance, I think part of the issue is that there's some -- let's say there's some kind of bug, you realize there's a problem. And you just need to fix something. And if you are super user and you [inaudible] look, I realize that there's some problem with this operating system, we have to open the hood, fix it up, and the problem would be gone. So if you are using Linux -- and that's what you see Google doing, right, they are massively modifying Linux and contributing to Linux, by the way. And I would be guessing that what happens is that they are doing something and they realize that there's one thing that doesn't work quite well and they fix it, and it's probably a lot of it is just bug fixing. And again, you want to -- if you are a super user, you want to have that control. If you're a Citibank you don't. Right? So I have probably what, like two minutes left? Zero minutes left? Minus two. So give me minus two plus minus two is equal to let's say somewhere around in the neighborhood of zero. So like two more minutes. So let me dwell on history for like exactly two minutes. So -- and hopefully it will be intrigue enough that you go actually go out and read the history part of the paper. I promise it's actually the most interesting part of the paper. It's really sharp, it's like before going to sleep. So open source is much older than open source. So in fact, the practice of -- the practice of developing software's public good predates proprietary software by [inaudible] even though the term open source is relatively new, it's like less than 25 years old, in fact less than 20 years old, their practice of public software, where you get the source and you can do whatever you want to, so true open source, is actually very, very old and predate proprietary software. In 1952, IBM entered computer market. They were the second mover. They were within the year of the first enterer. Less than a year. So at that time, the word software didn't exist. Obviously the word open source software didn't exist either. But software did exist. And at that time the software comes for free with the hardware and it is open sourcing, so whatever you want with it. And interesting at that time t amount of money -- so there's like this fake belief that expenditures on hardware exploded in the recent years. That doesn't seem to be true. So at that time, when first computers were launched in the '50s, the buyers of those computers would spend as much or more on software as they were spending on hardware. And IBM was very aware of it. Except that you wouldn't be buying software, you would be having a staff of programmers who are creating this software. But there's an expenditure on software which is huge. It was the same [inaudible] expenditures. So IBM realized it, and they wanted to increase efficiencies. We need more software, and it's incredibly expensive to do well. The users at that time were super users. You know, who should it be? It would be like one of the first five people in the -- five companies in the world to buy a computer, right? So those were really super users. And lowering the cost of programming was of course very important for IBM. This cheaper software is available software. People would be willing to pay more for hardware. So IBM wanted to encourage sharing of software. So, guess what? Within the first year of the launch of IBM first computer they created IBM share. A community of IBM users at that time the number of users you can sort of count on two hands for each system there was a separate share, at times there was only one system. So it was within the first year they create this IBM share. They get together, they share software. Before IBM creates share, those guys are already sharing, even without IBM facilitation. But IBM institutionalize it. And the -- here's what's interesting. And very insightful, I think. So two dozen organizations from members of share for IBM [inaudible] and at that time those people, they already understand the name of the game. They understand that sharing is its own reward. So you would think as an economist I would think, oh, who would want to create software, who would want to share, you know, I'll sit on my hands and try to sell. No, that's not the name of the game. So the contemporaries write that we shared -- so -- next slide. So the eyewitness accounts is that the real problem was not who would want to share, the real problem was everybody was first shoving their software at the throat of the other guy. And the reason is that those guys, even though the word network effect wasn't invented yet in 1955, they already understood it very clearly, they understood that by getting you to use my software, and I already developed a bunch of infrastructure complementary to this software, right, I was in this game for six months, if you would start using my software, you would obviously start improving it for your purposes, because you would like run into lots of problems. And then you would share it with me, of course. So I would directly benefit from you using my software. So they were shoving their software into each other's throats. And according to Grinwold [phonetic], United Aircraft browbeat share into accepting their assembler, threatening to leave the group if another proposal is accepted, right? So that is the early days of open source software. So the institutions of open source software completely fascinating predates open source software by several decades, and they evolve in most interesting ways. Thank you. >>: [inaudible] trying to shove this papers down everybody else's throat in [inaudible]. >> Michael Schwarz: Well, so in fact, I think the analogy is quite appropriate. So for -- quite appropriate, because I think the reason that Stallman felt that sort of free software as in free speech, rather than free beer, I think that's exactly what he meant, right? When you write a paper, your paper is in some sense not at all open source, but in some sense it is, so technology of science is a little different from technology of programming. But I think the way Stallman saw it, he saw software as kind of expression of human intellectual activity, an output of that, and he wanted other people to be able to build on that, to be able to create new stuff from that, and the price wasn't what worried him, he would have been probably fine if, you know, there is a reasonable system of payments there. But he didn't see how it would be possible to have a reasonable system of payments. But that's exactly what worked. He was like proprietary science versus open science, the science based on which other people could build, and he felt that it's imperative for the human endeavor for people to be able to build based on the software that came before them. And that's why computer programmers feel very strongly and emotionally about open source, even though, hey, they're not the ones who are paying the prize of software, they are happy when software is expensive. And the users like me don't feel passionate about open source at all. Though don't see moral issue there. Even though the users should be very happy about open source, hey, isn't it free, but it's the users who don't care and programmers do. And that's why it's free speech. Because for programs it's free speech but for users it's free beer. Thank you. [applause]