Someone recently asked me why I thought the vision I've been promoting around personal clouds can come to fruition in the face of Facebook's awesome power. With 900 million members and $18 billion in the bank, isn't Facebook becoming the one identity provider to rule them all? Won't everyone just use Facebook as their cloud? Shouldn't we all just get used to writing Facebook apps and forget about the business of personal clouds? These points (and others) are hard to argue with. Facebook seems like an unstoppable juggernaut.

And yet, we've seen such things before. AOL in the early 90's had that same feel to it. AOL was synonymous with going online. And yet something did upset AOL. A little thing called the Web. Note that I didn't say "Internet." The Internet was around in 1990, but no one on AOL cared about it. All you could do with the Internet was send mail (which AOL users could already do) and FTP stuff...whatever that was. There were lots of companies who had a presence on AOL. AOL was the gatekeeper to it's 2.5 million members and if you wanted access, you dealt with AOL.

And yet, the world moved from a centralized system (AOL) to a decentralized system¹ (the Web) in a few years. The problem with comparing the Internet to AOL in feature-set and functionality terms was that as a decentralized, open system, the Internet had tricks up its sleeve that AOL couldn't match. When the Web came along, AOL's fate was sealed. Now, Facebook has created a centralized system for social interaction. Compared to what you can do with Internet-thingies—like blogs, it looks pretty sexy. And still, I think it will eventually fall to a decentralized response.

We seem to yo-yo back and forth between centralized and decentralized online interaction models. This isn't surprising. Humans love centralized, hierarchical systems. They are easy to understand and predictable. They're tame. They're easily controlled. Consequently, we default to centralized when we build things.

Decentralized systems, on the other hand, are often described as wild and unpredictable. Decentralized systems are hard to understand and even more difficult to control. They defy our efforts to rein them in. But, for all their benefits, centralized systems have some big flaws. They are single points of failure, they limit freedom to participate and they don't scale well.

The (Eventual) Fall of Facebook

Let's focus on scale for a bit. As I said, scale is hard for centralized systems. See Geoffrey West's talk a TED for an entertaining presentation of the surprising math of cities (decentralized) and corporations (centralized) for a great explanation of this. Still, you might argue that with 900 million members, Facebook has scaled pretty well.

But scaling is more subtle than simply adding new users. Scaling involves providing the services that keep everyone happy. Let me give an example. Suppose you're responsible for innovation at a bank. You're trying to figure out how over the next decade your bank can increase revenue by providing new products and services to customers. Do your think that your boss will be satisfied with a "social media strategy?" Are you going to create significant new revenue for the bank by creating a new page on Facebook or tweeting better? Even a Facebook app isn't going to do much.

The problem is that Facebook's centralized architecture has intermediated the banks. Facebook has the relationship with the customer, not the bank. And Facebook's business model is to squeeze value out of that relationship for Facebook. There are things that banks will want to do that aren't in Facebook's interest. For example, Facebook is unlikely to enable features that make banks happy if those same features make the movie industry unhappy. A centralized system implies that decisions are made centrally for the organization's best interest. Even more difficult, not every bank wants to pursue the same strategy, so centralized systems almost always end up leaving more and more people dissatisfied as they scale.

Replacing Facebook

Facebook is primarily in the relationship business. But relationships can be decentralized. When the Internet's relationship infrastructure is decentralized more companies will be able to play a role in providing relationship services and those services will be more varied and customizable.

Again the case of AOL is a good example. AOL was good enough until it wasn't. What replaced it was a decentralized system that provided much the same thing without the restrictions of the centralized system that AOL built. At the heart of that decentralized system were two protocols: SMTP and HTML. All of a sudden, anyone could stand up a server and send email or share Web pages. People quickly went beyond the simple kinds of pages that the Web provided, using HTML as the vehicle for creating the rich range of online services that constitute the Web.

Besides those two protocols, there was DNS, or the Domain Name Service and the URLs that rode atop it. Most people don't realize that DNS wasn't in widespread use until the early 1990s. Following quickly on the heels of DNS was the invention of URLs, a system that provided universal naming for anything you could put online.

Protocols represent a way to do things that lets anyone participate and solve their own problems. Protocols are our primary tool in moving from centralized to decentralized systems. Consequently, people and companies who are unhappy with Facebook's offerings and service have a real choice for something else. The services Facebook offers can be replaced by a decentralized system based on protocols.

Will it happen? The history of commerce is littered with the memories of seemingly unstoppable giants. There's always something new under the sun. That history would also tell us not to look too closely for something that looks just like Facebook but is decentralized. That's rarely what happens. Disruption usually occurs in the weeds, where the incumbent isn't even trying to compete. The thing that undoes Facebook likely won't even be seen as a social network in the way we define those terms today.

Ways not Places

I've become a proponent of ways over places—that is, of protocols over Web sites. When it comes down to it, Facebook is a Web site—a great big, complicated Web site. The fact that it has a programming model (apps) and an API do nothing to change the fact that it is centralized and thus ultimately limited in scale and scope.

More people—especially geeks—need to understand that protocols are worth the effort they require. And let's not dodge that; decentralized systems are more work. They are harder to design, harder to build, and harder to bootstrap. But the forcing function, freedom, is powerful and eventually wins out.

Bonus Material

• Doc Searls has a great piece on the real story of Send that discusses the importance of protocols.
• Dave Winer has written We *can* do better than Facebook discussing why the seeming inevitability of a Facebook dominated world doesn't square with history.

1 Note that I really do men decentralized. A centralized system is one where a single agent makes a single decision, a distributed system is one where multiple agents make a single decisions, and a decentralized system is one where multiple agents each make their own decision.