Subject: Re: Tim's paradigm shift
From: Tim O'Reilly <>
Date: Wed, 7 Jul 2004 13:20:54 -0700

Sorry to be so long about responding to this thread.  I'm just catching  
up after being on vacation in the Galapagos.  (Very highly recommended,  

On Jun 29, 2004, at 3:27 AM, Matt Asay wrote:

> The problem I have with it today, just as I did then, is that it
> provides little normative direction.  It essentially says things like,
> "The way to succeed in the future is to build websites that everyone
> uses, because massive participation is a great barrier to entry."   
> Fine.
>  But how do I get there from here?

OK.  Here are some concrete examples:

1.  Let's rethink email and address books for the age of the internet.   
Social networking apps like orkut, friendster and linkedin are all  
hacks -- trying to recreate the metadata that's already implicit in all  
the email we send to each other.  Until I saw GMail, I was convinced  
that once Microsoft woke up to this possibility and built social  
networking into Outlook (extending the ideas they've already shown at  
Microsoft Research in Lili Cheng's  , they would lock up this  
opportunity.  So now, maybe Google is going to do it.  But a project  
like Mozilla or Chandler could also show some real leadership here --  
social networking could be implemented WAY more effectively by an open  
source email/IM client, that used a peer-to-peer address book to build  
up an unassailable lead in the form of a huge database of  
user-contributed information.  There are privacy issues here, but I  
believe that a little bit of inventive indirection could give a lot of  
functionality without revealing too much.

2.  Let's take the ideas of Nat Friedman's dashboard, and think how  
they extend out to internet scale.  But in any event, let's build a  
next generation Linux desktop in which the internet is seamlessly  
integrated, in which both the web and handheld devices are not second  
class citizens but a central part of the paradigm.  (Apple's non-open  
source iPod/iTunes can teach us a lot here about how to build what Dave  
Stutz calls "software above the level of a single device.")

But it's not just leveraging the "infoware-oriented", user-contributed  
database opportunities up the stack".  I also talked about the lessons  
of Dell.
Getting better at release engineering, building custom distributions,  
and live update seem to me to be core competencies.  Reinventing these  
things as internet-style competencies rather than enterprise software  
competencies (e.g. learning from eBay or rather than IBM  
or BEA) may be important positions to take.

There's also the opportunity to find an "intel inside" type of  
position.  MySQL may already be well on its way to this kind of  
position.  Dual licensing may allow this as an option for a widely used  
piece of infrastructure software, but I suspect that Intel Inside style  
lock in may more likely be found in various kinds of data.  Of course,  
if you have free software ideals, any source of lock-in may be seen as  
anathema.  But in that case, the normative direction is to make sure  
that you don't ignorantly give away your freedom (as we did with the  
DNS) because we don't understand the paradigm shift.

> I suppose I should architect my software/service/site such that it
> breeds add-ons, but this is no guarantee that anyone actually will
> choose to add on.  In other words, good architecture doesn't in any way
> ensure a good product/good sales/good IPO.  Maybe it makes these things
> more likely, but I'm not convinced.

There's no guarantee with anything.  Good architecture won't make your  
product sing if it doesn't meet a real need.  But I'll lay odds that if  
you hit a sweet spot in the market, the product that is architected for  
participation will win disproportionate market share and  
disproportionate customer stickiness.  That's the point of my  
illustration about MapQuest vs. the other innovative killer apps.  They  
don't have anything unique, where eBay, Amazon, do.

> And it doesn't explain why MySQL has been so successful.  They employ
> 100% of the code committers to MySQL - no architecture of participation
> (unless you count QA, and that does count).  They're open source in  
> some
> different sense of the word than the purists would like to use, but  
> it's
> a sense that is bringing them dollars.  More on this below.

They have neatly slotted themselves into a component of the overall  
architecture.  Intel didn't have a participatory architecture either.   
But by making a cool strategic bet to single source the 386, they were  
able to build a lock-in position within the overall open architecture  
of the PC.  Databases are so central to next gen applications that I  
think mysql has a similar opportunity.

The problem I think you're having, Matt, is that you are looking at my  
points as if they represent the whole universe of conditions for  
success.  They represent only one axis, one whose importance I think  
has been overlooked.  But far from the only driver of success.

Tim O'Reilly @ O'Reilly Media, Inc.
1005 Gravenstein Highway North, Sebastopol, CA 95472
707-827-7000 (company), (personal)