Notes from Tim O'Reilly's talk at Reboot 2003: reboot.dk Impressionistic transcript by Cory Doctorow doctorow@craphound.com 20JUN03 -- "The Open Source Paradigm Shift" Paradigm shift: Introduced in 1962. Used to describe the change from Ptolemaic to Copernican astronomy. At certain points in intellectual history, everything you know turns out to have been wrong. The last paradigm shift in IT: Mainframe to minis to PCs. They were Computer Companies: hardware and software. Apple is the last of these. IBM built "open source" hardware with the PC. Commodity hardware kicked Apple's ass. Increased commodification went on: Dell beat IBM and Compaq (because Compaq added bells and whistles, made a premium product). We're in the middle of another paradigm shift: Linux critic: Linux isn't user-friendly Linux geek: Linux will be better in the next rev They're both wrong: the apps that run on Linux are Google, Amazon, etc. Not shrinkwrapped apps, but new platforms. Amazon, Google aren't OSS, but they're built on OSS. Getting the source to Amazon would be pretty useless, in fact. Kurzweil: Inventors have to be interested in long-term trends. Inventions have to work int he world in which they're finished, not started. Licenses aren't the heart of OSS, they're just hygeine. What matters are three trends: 1. Software commoditization 2. User-customizable systems 3. Collaboration -- Commodity Amazon switched from DEC Alpha to Linux and saved an order of magnitude. Apache's existance means no one makes money on servers (Netscape couldn't sell browsers OR servers, MSFT had to give away IIS). Oracle will be in the same boat too. The Internet's architecture supports this through open standards and plug-and-play. Proprietary alternatives have to be free as in beer to survive. Vendors are acting like early PC vendors, like Compaq. IIS comes as part of a "server grade" version of Windows. This will break down and we'll see the end of this part of the software industry. Internet Application Platform: Intel commodity hardware, IP, Linux, Apache, MySQL, perl/Python/PHP -- Customization 1770: The Mechanical Turk. A hoax chess-playing machine. Babbage lost to it, and he knew it was a hoax, but he wanted to know if it was really possible. The secret of the Turk is that there was a man inside. Every app has a programmer inside. Take the programmer out of Amazon, Google, etc, and it stops working after a while. This is different from desktop software. This is customization at work. Google/Amazon/eBay are updated constantly, but MSFT only revs every 3 years. -- Collaboration Predates OSS. Computer networks provided the platform for collaboration. Unix's babel of hardware meant that you had to distribute source, not binaries. Usenet replaced tapes-in-the-mail. The leading OSS projects' key developers are spread all over the world. It's an ad-hocracy. OSS is the front -edge of a social phenomenon whose time has come. Power shifts from companies to individuals. Marc sold Macromind to Macromedia and lost control. But since OSS isn't officially owned, it can't be sold. OSS developers have implicit property rights. When Torvalds moves from Transmeta to OSS Research Labs, the heart of Linux moves with him. Collabnet teaches companies how to manage distributed development. OSS techniques work in proprietary software, too. MSFT acts like an OSS project, only in-house. ASP.NET came about because two developers at MSFT had 6 weeks to kill. It spread inside MSFT and got sucked into a lot of other projects. Shirky's "Listening to Napster:" You can build a big database by: 1. Paying people (Yahoo) 2. Getting volunteers (DMOZ) 3. Architect the system so that users' natural activity yeilds a database (Napster) That's the secret of what's happening on the Internet today. Every time you make a link, you contribute to Google. Amazon reviewers improve Amazon. This outstrips OSS projects for collaboration. -- The bizmodel for a commodity, data-rich app, collaborative, customizable world: * IBM Websphere: combine Linux/Apache and database with other foo and turn it into a big thing where no one knows what's free and what isn't and so everyone buys it and pays for support * OS X: Proprietary, free and added value * A commodity software platform, like a Dell for software * But: the lesson of Intel Inside is that there is some bit of proprietary value that can become a part of everything (.NET, Java, Digital Identity, search, etc.) * BIND: A monopoly in disguise. Made a fortune for NSI, not the guy who wrote it. * Sendmail/apache: revenue from services * Google/PayPal/Amazon The money doesn't come from selling software. People who don't know this have a paradigm failure. -- What keeps me up at night? * No one in OSS seems to care about Google/Amazon/eBay (I told Bezos off for pissing in the well when he sued BN.com) * Users don't own their data -- who cares about source when your data is locked in?