I noticed that when we provide a string that is not a IRI to a field that is expected to be an IRI then RDFLib generates an IRI using

e.g. for this JSON-LD (schema.org defines license as an IRI)

{
  "@context": "http://schema.org",
  "license": "notspecified"
}

After parsing and serializing back to turtle I get (path being the current path):

schema:license <file:///Users/vemonet/dev/notspecified> ;

It seems a bit like a hacky behavior to generate the IRI. Is this really needed? In which use-case is this actually relevant to have this behavior?

What could be a better behavior?

• Throw an error? Arguably the RDF we try to parse is faulty, and for reliability throwing an error would be better than silently altering the underlying data (there could be a strict=True param on the parser), which can cause many more problem down the road
• Use a named blank node instead of an IRI (I am not a big fan of blank nodes though...)
• Generates an IRI without using the absolute current working directory path

I can look into this and implement the changes if ok