Typed literals text

I am working on new text for the abstract data model, gtiven that I am not 
going to get it to Graham before tomorrow, the full version of the doc will 
not be ready for agenda considerations for this week.

Hence I include here my current fragement on literals; to give the chair an 
opportunity to consider whether we need to discuss any of the issues raised.

I also include an alternative that represents what I think I heard as the 
minority opinion at the Sept 13 telecon.

Note sections 3.2.1 3.2.2 are unchanged except for the treatment of null 
language tag, that follows the new xml treatment of using "" as the null 
language tag (I would particularly value DaveB's feedback on that change).

The new stuff is in 3.2.3.

The key philoshophical differences are:

+ are we doing *RDF* datatyping or *XSD* in RDF.
+ is the normal RDF environment datatype aware or not

As I see it, the WG preference is for RDF datatyping but not assuming any 
awareness of any actial datatypes (hence a preference for lexical forms and 
lang tags).

The minority opinion is that we are primarily doing XSD, and the target 
implementation knows them all. (Hence a preference for values, or 
pseudo-values).

Another argument that came up on Sept 13 was not over-zealously discarding 
user input. 

Concrete textual issues are:

+ lang tag - in or out
+ do we talk about the typed value here or only in the model theory
+ do we want to restrict to the canonical form

Jeremy