World-wide computer

tion was of limited impact when restricted to a local area, so machine manipulation of knowledge may blossom when it takes place on a global scale in a manner similar to the use of hypertext concepts in the Web. Today, the principal use of machines that read Web data is for search engines. These are notorious in their ineffectiveness—this is not due to any fault in their design, but because they must operate in a world designed almost exclusively for human consumption. A Web indexer has to read a page of hypertext and try to deduce the sorts of questions for which the page might provide an answer. Consider instead a Web in which, for example, the records of ownership of land, offers for sale, ownership of groups, endorsement of documents and individuals by bodies, and expressions of who owns what, who belongs to which groups, are all available in machinereadable form. This means that these documents are statements made in a computer language rather than a natural language, and that the statements, or assertions, are made in terms of things defined either in similar assertions or (as the system must be tied to reality in some way) by a human-readable definition of basic objects and operations. For example, it may be interesting to record that a person is a member of a company. The company may be defined by a statement in a registry of companies, employees may be defined by pointers to their Web pages, and the relationship of membership may be defined both in terms of its significance to people, and (for the computer) in terms of its logical properties and implications. The term assertion is appropriate, since it emphasizes that some party is asserting something: in general, it will be necessary always to be aware of who said what, if the results are to be trusted at all. Already we see the need in some Web pages to include incidental information (e.g., Tim Berners-Lee