In the WC3 document object model, a document is an abstract thing: an element with text, comments, attributes, and other elements nested within it.
In the semantic web, we deal with a set of "triples". Each triple is:
- a subject, the thing the triple is about, the id, the database primary key - a URI; and
- the predicate, the "verb", the "property", the "database column" - another URI; and
- the object, an atomic value or some URI.
OWL is to the semantic web as Schemas are to the W3C document object model. It documents what the various URIs mean and specify how they are used in a formal way that can be checked by a machine. A semantic web may or may not be valid with respect to the OWL that applies to it, just as a document may or may not be valid with respect to a schema.
RDF is to the semantic web as XML is to the DOM - it's a serialisation of a set of triples.
Of course, RDF is usually serialised as an XML documents ... but it's important to understand that RDF is not the same thing as "the XML serialisation of RDF".
Likewise, OWL can be serialised using OWL/XML, or (sorry about this) it can be expressed as RDF, which itself is usually serialised as XML.
