The recent launch of, the first application of the semantic web engine being developed by Danny Hillis’ new company, Metaweb, was written up by, among others, Esther Dyson, Tim O’Reilly, and Martin Heller, from whom I received an invitation to try Freebase. (Note: I don’t yet seem to have invitations that I can dispense.)

If you scan those articles and the blogospheric halo surrounding them, you’ll soon glean the essentials. Freebase is like Wikipedia in the sense that it’s an open data project. But where Wikipedia is a database of unstructured articles, Freebase is a database of categorized and related items. You can use it to add or edit items and, more ambitiously, to create or extend the categories themselves.

There’s been a lot of discussion about how this approach does or doesn’t match up with the W3C’s vision for the semantic web, and the suite of standards and technologies associated with it. I’ll leave that to the experts and simply reiterate one crucial point. The authors of the semantic web are going to be people, not machines. And people will only want to play the game if it’s easy, natural, and fun.

Early indications are that Freebase is going to be a whole lot of fun. In his walkthrough Tim O’Reilly calls it addictive, and explains why. Because the system thinks in terms of relationships among types of items, a single act of data entry can produce multiple outcomes.

Tim’s writeup gives a couple of examples of what that’s like. Here’s mine. I found a record for myself in the system, sourced from Wikipedia. I updated it to say that I’m the author of the book Practical Internet Groupware. Then I added that Tim O’Reilly was the editor of my book. That single edit altered the records on both ends of the author/editor relationship. My book’s record now showed Tim O’Reilly as its editor, and Tim’s record sprouted a Books Edited list that contained my book as its first item.

Nice. This is just a Hello World example, of course, but it has the feel of something that people will be able to understand, will want to use, and will enjoy in a social way.

A couple of years ago, I wrote a column entitled WinFS and social information management. It concluded like so:

Developers have always tried, and so far always failed, to define reusable objects that meet the needs of knowledge workers in the real world. Meanwhile, in the era of social computing, we’re learning to watch for the patterns that emerge as people interact in information-rich contexts, and then pave those cow paths. The first WinFS-aware applications, which will be personal information managers with hooks for sharing and synchronization, won’t align with this strategy.

These WinFS applications will, however, enable you to pave your own cow paths, for example by storing and reusing queries. Nobody can know how people will ultimately want to share these contexts among WinFS clients in a peer-to-peer fashion, on WinFS servers when they emerge, and on the global XML Web. So I hope Microsoft will come to see WinFS not only as a platform for developers, but also as an environment in which users can do simple things that yield powerful social effects.

Nowadays a lot of folks say WinFS was doomed from the start and should never have been attempted. I didn’t think that then and don’t now. I did, clearly, wish that WinFS had been part of a strategy of cooperation with the cloud. And I’d still like to see some version of that scenario play out.