Not every string will be an entity, otherwise it would be turtle-entities all the way down :)
We need entities whenever there is some thing that should have identity. 

The naming conventions specified in linked data best practices help to remind implementers whether the predicate should have a literal or a resource as its object.  Datatype properties should be nouns, Object properties should be verbs.  Bibframe is currently very inconsistent in its naming, and tidying this up would be a great benefit.

Which is not to disagree that documentation and examples should be available, comprehensive, consistent and clear.


On Thu, May 14, 2015 at 9:13 AM, Wallis,Richard <[log in to unmask]> wrote:
On 14 May 2015, at 16:54, Robert Sanderson <[log in to unmask]> wrote:

At which point a consistent model, that is easier to construct, parse and render (e.g. always a resource, sometimes a blank node) is clearly better.

Which brings me back to my vocabulary consistency point - will all strings be entities - or at least pseudo entities using blank nodes?

If not, and some properties will be defined as being string properties; that is fine but it is indicative of the basic documentation and examples needing to be very clear on when you have a string and when you have [at least] a rdfs:label wrapped in a blank node.  Remembering a natural understanding of RDF and its principles will not be an attribute of most of the folks applying whatever we end up specifying.

Rob Sanderson
Information Standards Advocate
Digital Library Systems and Services
Stanford, CA 94305