Sitting in a conference: OSCon -- Tim O'Reilly Keynote
Here at hardcore geek central, the O'Reilly Open Source Conference. Microsoft is sponsoring lunches -- Tim O'Reilly had a nice quip from Slashdot: there will a click-wrap license on the lunches.
Tim starts by asking us to think of a paradigm shift, as though we haven't heard that idea enough. I think it's over used, this shift idea. While it comes from Thomas Kuhn's book, The Structure of Scientific Revolutions, it has been trivialized.
Tim says think of the ways the rules changed with the IBM PC. First IBM introduced the PC in 1982, based on standard parts and released the spec, changing all the rules of the computer industry. Before that you had to build your own integrated OS and hardware (not quite -- UNIX demonstrates otherwise). Says Apple is the last survivor of the proprietary model -- what about Darwin? -- but IBM/Microsoft still own 95 percent of the market (leaving, by extension to this statement, Linux with all of two percent, which is low).
Michael Dell's innovation was making a better process for building standardized PCs. This is a shift in power in the personal computer space from Apple to IBM to Dell.... The branding of components, like "Intel Inside" will point to the way to a successful Open Source strategy.
He moves on to the decoupling of hardware from software, leaving the market open for Microsoft to become the most important company in the PC industry.
On to Linux: All these people argue about Linux. It's hard to use, geeks make the claim that Open Office getting better, Gnome is great -- but all this is lost on the front end. Killer apps, like Google, Amazon, PayPal and Yahoo, run on Linux. So, what is the meaning of the shift that is going on?
Several things are wrong with this picture from the Open Source perspective. Most companies that use Open Source are fiercely proprietary. Google doesn't expose its source, nor does Amazon; and there is little difference in the code -- it's in the data and customer critical mass they've accumulated (very 1999 thinking, which is true, but not financeable anymore.) William Gibson said "The future is here, it's just not evenly distributed yet." And that's where we are today.
In the same way that the PC led to open hardware architectures (even at Apple, where memory and PCI and so forth are standardize now -- yes I am typing on a Mac). Now the move is to open softwarte architectures. The era of proprietary APIs as a key to advantage is ending, Tim says. He's suggesting that the Intel Inside and Cisco Inside approach ends up being the most powerful approach -- but this is integrated hw/sw and in a way, I think, there aren't many steps from the old paradigm here. It is a difference in what is collected around remaining defensible proprietary architectures.
Software as Commidity: Software is driving down margins and increasing competition and Linux delivers order of magnitude cost savings. Apache has eliminated Web serving as a business model and MySQL is doing the same in databases -- though he says it's still being debated.
In this environment, software is deeply componentized and no one can tell when an OS on a server is changed. The proprietary options are bundled with added value components to keep a price advantage.
Customizability at Work: Upgrade cycles are maniacally accelerated, even to daily releases. You need dynamic languages to prevent the need to recompile. This is why Perl, PHP and Python -- higher level languages -- are the answer and are being adopted everywhere.
Interesting idea: There are people inside the software we use (Tim takes this from the story of Van Kempelen's Mechanical Turk) and if the folks at Amazon or Google stopped working, the services would fail. In fact, this is interesting in terms of employment and the value of labor -- but I'll deal with that at another time.
Communications is critical: The distributed teams that have built Open Source software are the new work paradigm. (I agree, see my story about InnovationWorld here). Network-enabled collaboration is breaking out everywhere.
He goes on to the rise of ASP.net at Microsoft. The creators were told they couldn't make ASP XML aware because it would break backward compatibility by the bosses. Went on vacation and hacked on a version -- when they cam back to work and people said they wanted pieces. Bill Gates calls a half year later and, after reviewing, made it a flagship product for Microsoft. Networking is happening everywhere. Tim says "If it is happening inside Microsoft, it is happening everywhere." I think this is underestimating the intelligence of Microsoft. They've been doing this for decades, albeit they force people into labyrintine buildings to work.
Collaboration at the data layer -- if you set up a network that is a byproduct of selfish action, it yields a huge result. This is true, but moving from uncoordinated to coordinated action is more difficult than Tim is letting on. Google is a reflection of this huge collective work, Tim says. But Google is famously secretive about what they do, adding a secret sauce to all the searches. We are all contributing, as Tim says, but we're not giving away our effort to Google. Google is mapping our usage, so it isn't that we're volunteering, but that we're being described -- the collective intelligence of the world is finally being described. In this context, architectures become critical, a kind of public space like physical world architects work with, an intellectual act about how we should flow together in the dataspace.
So, Tim says, we're just discovering the business model. It's about bundling. When is someone going to come along and really figure out the Dell model for Open Source, Tim wonders? (I think the issue is not that large -- how will small pieces of function be assembled.) Tim says "It's about building Linux as a process." Yes, absolutely.
There are many "chokepoints" that are possible "Intel Inside" plays: J2EE, .Net, Digital Identity, and the Web services business will be aggregated into packages people subscribe to. It's the updated software business we've been talking about for a decade, I think. Not paradigm shifting, but realizing a long process of convenience from the packaged to the networked application.
So, it's about services delivered to users, not "professional services." Tim says UUNet is the greatest Open Source business yet, and so is BIND, the DNS system is operated on (and which is a pain in the ass, if you ask me). Sendmail and Apache are hosting plays and look to the networked apps like PayPal and Google as being indicative of the changes ahead.
"I believe we're building an Internet operating system," Tim says. Here are the parts of the puzzle:
* P2P and ad-hoc networking
* Wireless
* Social Software
* Cell phones and mobile devices
Increase standardization while expanding user's ability to customize their experience. Marc Canter would be hooting with delight at this statement. Developers need to leverage the collaborative opportunity.
Tim says we should watch the "Alpha Geeks" as hackers prefigure future movements. Two examples: Screen-scraping described web services; wireless community networks predicted Wi-Fi market adoption.
Open Source needs to be rethought in terms of proprietary components.; Google and Amazon APIs are available to anyone, but proprietary: Who owns the data relationships?
Open Source is a bill of rights for software developers. (This is a mistake -- it must be a bill of rights for the end user.)
"We have to think more about where it is all going. Where is this paradigm going to take us?" Tim exhorts. History is where the answers lie, not in total breaks with the past, if you ask me.
Posted by Mitch Ratcliffe at July 9, 2003 09:04 AM | TrackBack