Re: RDF Model Theory Working Draft: Comment

On Sun, 7 Oct 2001, Graham Klyne wrote:
> At 03:46 PM 10/6/01 -0400, Arjun Ray wrote:

>> The problem is this: If an RDF graph has "one node for each uriref,
>> bNode or literal identifier", then the general representation of a
>> statement will need three nodes - not two - because all three of
>> the triple's terms are *nodal*: [...]

> I think the answers are all in the model theory:

I'm sorry, I don't see this.  I'm not questioning the internal
consistency of the formal model (at least, not yet!).  I'm questioning
whether Fig 11 from the RDFMS spec is a correct picture.  If I have a
triple (foo bar baz) and if all three terms - foo, bar and baz - are
distinct URIrefs, then the plain meaning of the statement

: the correspondence between an ntripleDoc and an RDF graph is that
: the graph has one node for each uriref, bNode or literal identifier
: in the document

would lead me to expect that any graph-like picture should exhibit
three nodes, one each for foo, bar and baz, because each one of them
is a 'uriref, bNode or literal identifier'.  Instead, there are only
two nodes - for foo and baz - and the label I would have expected to
see in a node (bar) is on an arc instead.  So, what happened to the
node labelled by bar?  Why isn't it in the picture?  Did some two
dimensional figure (a triangle perhaps?) get flattened into a line?;)

> - A URI that labels an arc may or may not also label a node.

This does not seem to be a correct statement.  In 'the graph has one
node for each uriref, bNode or literal identifier', does 'for each'
mean 'not all'?

> In all this, there is no requirement for an arc to be a node;
> indeed that would be a contradiction of the syntactic definition
> of the graph.  A node labelled by a URI and an ARC labelled by the
> same URI are quite distinct in the domain of interpretation.

But what happened to the node labelled by the URI?  Where did it go in
Fig 11?

I contend that whatever the correct pictorial representation may be,
Fig 11 is *not* it, and consequently the MT draft should be amended to
describe a correct picture instead.

Once again, if 's', 'p' and 'o' in (s p o) are all urirefs, then there
should be three nodes, labeled 's', 'p', and 'o'.  Where is the
'p' node?  (And, of course, what are the relevant arcs to connect
them?)


Arjun

Received on Sunday, 7 October 2001 11:01:42 UTC