Imagine the set of all knowledge as a giant xml document. Now a query is simply an XPath location to the document fragment that holds the answer to the query. The document can be initially undefined, filled in only as people query it. btw, XML Fragments is a work in progress.
Think about it another way. There is some unknown function which is all knowledge. An operator acts on this function and out pops matching values; the answers. Very quantum mechanics. My point being -- the original document needn't exist. It's just a way of thinking.