It seems so simple...

...deployment requires more complexity.

How do we identify people (or documents) on a world-wide scale?
What do machines need to know about people? agents?
What is a document? A book? Hamlet the 'abstract work'? Expressions of it? Printed publications? Actual physical books? Publisher vs library perspectives...
RDF does not guarantee good modeling!
In a cross-domain world-wide sytem, "goodness" of modeling remain context-dependent