Re: So now we have tidy literals...

On Mon, 2002-10-14 at 07:42, Graham Klyne wrote:
> 
> At 08:30 AM 10/14/02 +0100, Brian McBride wrote:
> 
> >At 16:49 11/10/2002 +0100, Graham Klyne wrote:
> >
> >>Now that we have tidy literals, do we actually agree what (tidy) kind of 
> >>thing they actually denote, so we can say something sensible in the 
> >>concepts document?
> >>
> >>I.e., in:
> >>
> >>   Jenny age "10" .
> >>
> >>is there anything to say about what the "10" actually denotes?

Well, it's identical to what "10" denotes anywhere else in
n-triples, and it's distinct from what "010" denotes,
and from what "abc" denotes, and so on.

> >> From past discussion, I'm expecting that the answer will be that a 
> >> literal denotes a composite value consisting of a Uniocode string, a 
> >> language code and an XML flag, or something of that kind.  That would 
> >> tally with the current abstract syntax description [1].
> >
> >Right, though DanC has been suggesting we consider that we two types of 
> >literals, each a pair of the literal and the string, one is a bare literal 
> >and the other is an xml literal.
> 
> Yes... I think I came closer to that in the tentative text, which you did 
> not quote.  But mainly, I wanted to make sure we're all facing the same 
> direction now ;-)

I'm confused by the above.

It seems to me that the class of Literals is a sort of union:

	1.a strings, resulting from
		<title>abc</title>
	where no xml:lang dominates the <title> propElt

	1.b lang-strings, resulting from
		<title xml:lang="en">abc</title>

	2.a XML infoset thingies, resulting from
		<title rdf:parseType="Literal">some <em>very</em> good
			stuff</title>
	[some text earlier in a thread said that this was a
	unicode string; I wouldn't say that; it can be serialized as
	a unicode string, as we do in n-triples. But that doesn't
	make it a string]

	2.b XML infoset thingies, with lang, resulting from
		<title xml:lang="en"
		rdf:parseType="Literal">some <em>very</em> good
			stuff</title>

	3.a datatype values, resulting from
		<date rdf:datatype="&xsd;date">2000-10-12</date>
	These consist of an absolute URI reference
	and a unicode string.

I dunno if we have 3.b lang-datatype-values. I hope not.

Nor do I know if datatype values of type &xsd;string are
identical to normal 1.a strings. It seems best, for users,
if they are, but it's sort of an ugly special case.

Test case:

	<rdf:Description rdf:about="#something">
	  <p1>abc</p1>
	  <p2 rdf:datatype="&xsd;string">abc</p2>
	</rdf:Description>

entails?

	<rdf:Description rdf:about="#something">
	  <p1 rdf:nodeID="X" />
	  <p2 rdf:nodeID="X" />
	</rdf:Description>

(please add that to the test collection, Jeremy/Jan/et. al.)

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Monday, 14 October 2002 10:34:28 UTC