Jeni's Musings

I think this is a delicate balancing act. While non-literal resources are generally more useful than literals, trying to eliminate literals from a dataset can end up fairly horrible. Where does it end?

<#me>
  foaf:name
    [ a foaf:Name ;
      foaf:component
        [ a foaf:GivenNameComponent
          foaf:spelling
            ( <http://dbpedia.org/resource/T> 
              <http://dbpedia.org/resource/O>
              <http://dbpedia.org/resource/B>
              <http://dbpedia.org/resource/Y> ) 
        ],
        [ a foaf:GivenNameComponent
          # ...
        ] ,  
        [ a foaf:SurnameComponent
          # ...
        ]
    ] .

That way madness lies. Vocab designers need to decide at precisely where it becomes useful to use literals, and indicate this appropriately in their schemas (using rdfs:range).

Where to do this depends a lot on exactly what they hope to achieve from their vocabulary. If you want to model, say, products for sale, colour may be a small, incidental feature, and you can get away with using literals. If you are writing a tool to convert, say, CSS to RDF for a smart agent to reason with and determine, say, whether a particular design is accessible to the colour blind, then colours are going to be much more important to you, and there may be benefits in minting URIs for them.

And, for what it's worth...
http://ontologi.es/colour/F0A000