Tim O’Reilly opens this year’s conference with this talk. He begins with
the definition of paradigm shift: a change in world view that
calls everything you know into question. This is from
The Structure of Scientific Revolutions by Thomas Kuhn.
First, we had a hardware paradigm shift: commodity hardware with an open
architecture with the IBM PC. This resulted in killing off old proprietary
hardware companies: Digital, Data General, Prime. Apple is the last
remaining company in this space. We then had a software paradigm shift (IBM
let a tiny Seattle company provide the operating system for their PCs).
Open source and ease of use. Is it too hard to use Linux or BSD? No,
there are lots of user-friendly applications on those platforms:
Yahoo. But these applications aren’t
open source themselves, even though they are created with open source tools.
One of Tim’s favorite quotes: “The future is here, it’s just not evenly
distributed yet” from William Gibson. I think I’ve heard him reference this
in half a dozen talks.
Now we’re in the Internet Paradigm Shift. As software
architecture gets more open, it becomes a commodity, so competitive
advantage must move “up the stack” to services at a higher level:
information applications”. Now lock in is based on data nd customer
The open source community spends a lot of time talking about licenses,
but Tim wants to go beyond that to the Three C’s: Commoditization of
software; user-Customizable systems and architectures; and network-enabled
Commoditization of software: open source increases competition and drives
down margins. Using Linux on Intel can provide an order of magnitude
savings over proprietary software. Apache led to selling web server
software not being a profitable area, and MySQL threatens to do the same for
databases. The open architecture makes “plug-compatible” software the norm.
(I know I’m demanding standard interfaces, even in proprietary software. I
avoid using proprietary functions in databases, so Oracle is going to have
to do the basics better than MySQL to keep my business; it’s not going to do
it with unique functionality.)
Customizability. If you have commodity components for your
infrastructure, you want to build software “for use” rather than “for sale”.
Internet-era applications are on daily release cycles to enhance their use,
while Microsoft is now on a three-year release cycle.
Tim discusses “Von Kempelen’s Mechanical Turk”, a chess playing machine
from the 18th and early 19th centuries. Of course, it was a hoax: there was
a man in the machine playing the game. (Although it was clearly a hoax, it
probably provided some inspiration to Charles Babbage, who saw it, to create real
computing machines.) Internet applications have “men in the machine”. If
Microsoft closed down, Microsoft Office would continue to work. If everyone
left Amazon.com, that application wouldn’t last a day.
Network-enabled Collaboration. Everyone is interested in fancy cutting
edge collaboration tools, but Usenet is the real mother of open source.
Lots of important software was developed by internationally distributed ad
hoc teams of developers who never even met each other. Users help to build
the applications, too. Tim calls this “the adhocracy”. Tim notes that
Microsoft’s ASP.net is exhibiting open source software-like behavior. Two
Microsoft developers wanted to make ASP (active server pages) XML aware, but
their supervisors didn’t think it would work. They fiddled with it on their
own, and other people in Microsoft got interested, too. Half a year later,
Bill Gates called them and said he’d like to see their project. He had a
lot of suggestions, and then announced that it was a key initiative at
Data collaboration. Clay Shirky’s “Listening to Napster” says music
swappers build a network as a byproduct of their own self interest.
Google’s raw data is millions of independent links, not something it created
itself. Similarly, everyone who writes a review on Amazon (or even buys a
book) has contributed to Amazon’s valuable store of data.
Business model thoughts. Tim compares IBM WebSphere and Mac OS X to
Compaq. They’re very successful, but the commoditization of so much of the
foundation they do means someone is going to come along and do it better and
cheaper (the Dell model). Now we’re moving to web technologies like LAMP,
J2EE and .Net to become the “Intel Inside” of the internet. But the
platform is web services themselves, not the tools that provide them,
because they will only work if they interoperate. These things will be
aggregated and sold by subscription, not by the piece.
Hidden service business models in OSS. Now everyone is looking at
“professional services”, but services for end users. UUnet (the huge ISP)
is the greatest open source business success to date, not Red Hat. Paul
Vixie created BIND which provided domain name services, but never made money
from it: Network Solutions did, by providing root services for it as a
monopoly. Sendmail and Apache aren’t making much money for the software
developers, but they make a lot for internet service providers who sell the
use of those tools, not the tools.
Tim tells us to embrace the new paradigm. Use commodity
software to drive down prices for users. Provide customization capability:
plug-in standards-compliant compnents, extensible architecture, scripting
support. Look for hidden service business models, and develop collaborative
development and participatory interfaces.
Watch the Alpha Geeks. These hackers are first to exploit new
technologies, then entrepreneurs, then large players. Examples: screen
scraping showed that web services would have to happen; community wireless
showed that we would have Wi-Fi hotspots.
Closing quote from Ray Kurzweil: “I’m an inventor. I became interested in long term
trends because an invention has to make sense in the world in which it is
finished, not the world in which it is started.”