Not all data relations in RDF represent straight binary connections between resource and object value. Some data values, such as measurement, have both a value and additional information that determines how you treat that value. In the following RDF/XML:
the statement is ambiguous because we don't know exactly what 18 means. Is it 18 days? Months? Hours? Did a person identified by the number 18 edit it?
To represent more structured data, you can include the additional information directly in the value:
However, this type of intelligent data then requires that systems know enough to split the value from its qualifier, and this goes beyond what should be required of RDF parsers and processors. Instead, you could define a second vocabulary element to capture the qualifier, such as:
This works, but unfortunately, there is a disconnect between the value and the unit because the two are only indirectly related based on their relationship with the resource. So the syntax is then refined, which is where rdf:value enters the picture. When dealing with structured data, the rdf:value predicate includes the actual value of the structure?it provides a signal to the processor that the data itself is included in this field, and all other members of the structure are qualifiers and additional information about the structure.
Redefining the data would then result in:
<pstcn:lastEdited rdf:parseType="Resource"> <rdf:value>18</rdf:value> <pstcn:lastEditedUnit>day</pstcn:lastEditedUnit> </pstcn:lastEdited>
Now, not only do we know that we're dealing with structured data, we know what the actual value, the kernel of the data so to speak, is by the use of rdf:value. You could use your own predicate, but rdf:value is global in scope?it crosses all RDF vocabularies?making its use much more attractive if you're concerned about combining your vocabulary data with other data.