Maybe "datatype" should be in the RDF Graph explicitly

[HughGlaser]

I think this got lost in the discussion about language tags (which has also got very complicated). There may be a stronger case for taking the datatype out of the Literal and attaching it as a property to the Literal, or at least fewer of the negatives that apply to language tags.

It could (like lifting language tags) be simply done now as common practice. But would require SPARQL to take cognisance when doing sorting, etc..

I would find it much easier to do the RDF creation and consumption his way - anyone else agree?

[draggett]

It would be consistent with embracing property graphs where nodes and links can have properties. In other words, the data type for a literal node is expressed as a property of the node. However, this could also be expressed as a higher level framework that allows for properties, n-ary terms and so forth. This would hide any special treatment of data types and language tags at the RDF core.