Re: [core] Link target attributes in CoRAL (was: Review of CoRAL)

[Christian Amsüss <[e-mail]>]

On Sat, Nov 10, 2018 at 10:00:59AM +0100, Klaus Hartke wrote:
> Literals as link targets indeed seem to need some fine-tuning. Besides
> language tags, I can also imagine other kinds of information attached
> to a literal, such as a unit:

RDF has data types, of which many AFAICT could be mapped to non-string
literals (making their normalization issues go away like CIRIs do away
with IRI normalization issues).

Units I'd prefer to keep per-resource, but we can probably come up with
a way to make them into data types for RDF conversion too if we want.

> This may be read to imply that there is a resource identified by the
> integer 21, which has the unit degree Celsius and contradictorily also
> the unit kilogram...

The mind set to lead us out of this is to treat literals more like blank
nodes with a value -- two blank nodes are only identical if they are the
same null thing in the syntax tree, and literal nodes don't become
identical just by virtue of having the same literal value. (RDF might
treat them as identical if their value *and* all their attributes like
language tag and data type are).

CIRI references are not identical either if they have the same array
representation, but are identical if they resolve to the same full
CIRI (and as with RDF, some applications might introduce additional
equivalencies, eg. from protocol-negotiation).


But yes, this does need thinking over, and probably some text explains
why it's justified.

Christian

-- 
To use raw power is to make yourself infinitely vulnerable to greater powers.
  -- Bene Gesserit axiom