Dave Winer tries to answer why he is so vested in OPML.  Though he didn't really answer the question, he paints a picture that I think is becoming more and more interesting (and possible):

"I think the answer is to put the tools for constructing data formats into the hands of users. We've been going about it all wrong, coming up with straight-jackets for users, and expecting them to conform to some set of rules that make no sense about how the characters should be encoded, when it's ideas and relationships to other ideas that we want to make it easy for them to express. In other words, we don't know how to capture this data that we want so much, so create a tool for expressing data that at least some people can use, and let them use it, and let them make it public, and then see what kind of crawlers people write, and then learn, and iterate, and try again and again until we're closer to having the information we want at our fingertips, to quote another visionary from another time."

I'm not sure I see how OPML becomes the enabler of this vision, but it's a worthy vision nonetheless.  I really like the idea that we begin to express ourselves in terms of data and that the machines then learn to interact with us based on that expression.  Dave, are you suggesting that data is a new human language?