Re: #rdfms-literals-as-resources in scope?

On Thursday, July 5, 2001, at 08:27  AM, Brian McBride wrote:

>   o Does the formal model described in M&S have a distinguished
>     representation for literals.

I assume you mean abstract syntax, not formal model.

[...]
> If the WG decides YES on the first question, then to represent literals
> as resources in the abstract model would be a change.

Err, this is the point I'm not sure on. Would it not be possible 
to admit that Literals are Resources, define a set of URIs for 
them, but allow the abstract syntax to contain both forms 
(literal and resource)? This would not be a change to the 
abstract syntax, but merely some useful information to RDF 
users. Similar to the bit in the RDFS spec that says "Remember 
when we talked about the class of Literals? Well that has this 
URI...".

This is only one possibility -- I'm not necessarily sure that it 
is the best route.

> I have seen that several people think that it does.  When Aaron says:
>
>   Yes, there is a set of things in M&S called "Literals". Whether this
>   is a side-effect of the XML syntax, or of the abstract syntax is
>   not clear to me.
>
> I take it that he is undecided.  It would be helpful Aaron if you could
> come to view on this before the telecon, as I would like the WG 
> to decide
> this question.

I have come to a decision:

Since the M&S states that:

	All things being described by RDF expressions are called resources.

and (possibly among others) the xml:lang syntax allows an RDF 
expression to describe a Literal, then Literals must be 
Resources. Furthermore, since M&S states:

	Resources are always named by URIs plus optional anchor ids (see
	[URI]). Anything can have a URI; ] the extensibility of URIs 
allows the 	introduction of identifiers for any entity 
imaginable.

And I can clearly imagine Literals, then they must be RDF 
Resources. Since the spec is unclear on this point, I think the 
answer is no, Literals do not have a distinguished 
representation in the spec. Anything else would seem to cause a 
contradiction. I think you'd agree that it's easier to make 
Literals into Resources than redefine Resources as everything 
except Literals.

> The second question then is:
>
>   Is there a compelling reason why we should make this change?
>
> Compelling reasons might include - the current spec is unimplementable,
> is not being implemented, contradicts some other part of M&S or
> the change is needed to resolve some other issue.

Here's a starting point for some reasons:

  - The spec would be contradictory (see above)
  - Literals have many resource-like properties (language, type, etc.)
  - The "data:" scheme seems very close to a set of URIs for literals
  - Datatyping, query and other future directions in RDF may 
necessitate us to treat literals as resources.

Thanks to Dan Brickley for some of these ideas, although this 
does not mean he agrees with my position.

> Aaron I believe is suggesting that since literals have identity, they
> satisfy the only requirement to be a resource.  It would thus simplify
> both the abstract model and implementations to remove them as a
> distingished entity from the abstract model.

That is true.

--
       "Aaron Swartz"      |              The Semantic Web
  <mailto:me@aaronsw.com>  |  <http://logicerror.com/semanticWeb-long>
<http://www.aaronsw.com/> |        i'm working to make it happen

Received on Thursday, 5 July 2001 18:24:04 UTC