Summary

  • W3C's RDF Model and Syntax became a Recommendation on 22 February 1999; W3C announced it two days later. It described resources with subject, predicate and object statements.
  • The 1999 design identified resources by URI and tied each property to a namespace. XML carried the statements, while the named vocabulary told another program what a property was meant to say.

The early Web could show you a page about a book. Asking a program to find every book by the same creator was a different job. One page might say “author,” another “creator,” and a third might place a name next to a picture with no label at all. A librarian would ask for a catalog card. A machine needed the card's fields to mean something after it crossed a network.

The W3C Recommendation of 22 February 1999 called its answer the Resource Description Framework. W3C used a library catalog to explain metadata: information describing another resource. In the standard's model, a resource could be a page, a piece of a page, a whole site or even a printed book. A URI named it. That first step mattered. The catalog card had to say which thing it described before anyone argued over its title.

Next came the small sentence at RDF's center: subject, predicate, object. The subject is the resource, the predicate a named property, and the object its value. A value can be text or another resource. The Recommendation's own example says that Ora Lassila is the Creator of a named W3C home page. The page URL is the subject; Creator is the property; “Ora Lassila” is the value. A machine can carry that sentence as a structured statement instead of guessing from where the name sits on screen.

The original specification offered an XML form for exchanging these statements. An rdf:Description could name a resource once and hold several properties about it. It also allowed a regular serialization and a shorter one. The arrangement of angle brackets could change while the underlying model of statements stayed the same. In the catalog room, two clerks could stack the cards differently and still record the same assertion.

Then the librarian interrupts: “Creator of what?” The word might identify the writer of a page, the maker of a photograph or the person who compiled a collection. The 1999 Recommendation tackled this with namespaces. Each predicate belonged to a named vocabulary, or schema, where its intended meaning could be found. One description could draw properties from several vocabularies. The property name alone was a drawer label; its namespace told the next desk which dictionary had printed it.

W3C announced the Recommendation on 24 February 1999. Its contemporary examples ranged from discovery and digital-library catalogs to content ratings and privacy preferences. Those examples showed where a common description mechanism could travel. URIs, statements and vocabulary names supplied its working parts. The Web had acquired a way to send a catalog card to another program with the drawer labels attached.