- From: <Patrick.Stickler@nokia.com>
- Date: Tue, 8 Jul 2003 11:36:59 +0300
- To: <jbekaert@lanl.gov>, <www-rdf-interest@w3.org>
> -----Original Message----- > From: ext Jeroen Bekaert [mailto:jbekaert@lanl.gov] > Sent: 08 July, 2003 01:42 > To: www-rdf-interest@w3.org > Subject: URIrefs and granularity - cont. > > > > Hi All, > > 2 weeks ago, I have posted the message included below. Until > now, there was > no reply to this message. Therefore, I repost the message, > hoping that > someone will give us his/her (brief) opinion about this issue. > > Thanks in advance > jeroen bekaert > Los Alamos National Laboratory > > ----------------------------------------> > > Hi All, > > in addition to my previous emails concerning the removal of > the aboutEach > attribute in the current RDF spec., I have the following > questions I would > like to discuss with you. > > Suppose I have a statement consisting of: > 1. a subject, identified by: > http://foo/foo.xml#xpointer(Item[position() = 4 > or position() = 5]) > 2. a property: dc:author > 3. an object: "John Doe" > > Then the RDF/XML syntax looks as follows: > > <?xml version="1.0"?> > <rdf:RDF xmlns:rdf=http://www.w3.org/1999/02/22-rdf-syntax-ns# > xmlns:dc="http://purl.org/dc/elements/1.1/"> > <rdf:Description > rdf:about=http://foo/foo.xml#xpointer(Item[position() = 4 > or position() = 5])"/> > <dc:author>John Doe</dc:author> > </rdf:Description> > </rdf:RDF> > > As you might notice, the subject resource-identifier refers to 2 > subcomponents of the containing resource, namely an Item with > position() = 4 > and an Item with position() =5. This way, our proprerty is > attached to both > the subcomponents of the resource http://foo/foo.xml. Or rather, the URI denotes a fragment of the XML instance, a single thing, and that fragment has as its components, the two subcomponents indicated by the XPointer. > The > extent of the > subcomponent is defined by the fragment identifier in > conjunction with the > content type of the containing resource (i.e., text/xml). As > such, in our > case, the fragment identifier is an XPointer. > > So, in my opinion, according to the technical spec. of RDF, > this statement > is perfectly valid. It is. > However, I do have some concerns about the sematic > meaning of it. The above-mentioned example is stating the following: > (UNION of (http://foo/foo.xml#Item[4], > http://foo/foo.xml#Item[5])) has a > property dc:creator with value "John Doe". > > Now, one could argue if it also follows that: > http://foo/foo.xml#Item[4] has property dc:creator with value "John Doe" > http://foo/foo.xml#Item[5] has property dc:creator with value "John Doe" And if one chose to argue that, one could use OWL or a similar language to express arguments in the form of axioms/rules. E.g. <http://foo/foo.xml#xpointer(Item[position() = 4 or position() = 5])> a rdf:Bag ; rdf:li <http://foo/foo.xml#Item[4]> ; rdf:li <http://foo/foo.xml#Item[5]> ; { ?bag rdf:li ?item . ?bag dc:creator ?x . } log:implies { ?item dc:creator ?x . } Of course, I'd argue that such a rule would not always be valid. Since the creator (author) of a collection of items is not necessarily the creator (author) of each individual item. I.e. the person that wrote #Item[4] need not be the same as the one that wrote #Item[5] and neither need be the same as the one that arranged them together as they occur in foo.xml. So the statements that are made about the fragment do not necessarily apply to any identifiable entities within that fragment. It depends on your ontology, its underlying metaphysics, and what you are hoping to accomplish. > In a more general case, one could come up with the following example: > Let's assume we have a subject resource-identifier pointing to an XML file. > 1. a subject: http://foo/foo.xml > 2. a property: dc:author > 3. an object: "John Doe" > > Then the RDF/XML syntax looks as follows: > > <?xml version="1.0"?> > <rdf:RDF xmlns:rdf=http://www.w3.org/1999/02/22-rdf-syntax-ns# > xmlns:dc="http://purl.org/dc/elements/1.1/"> > <rdf:Description rdf:about=http://foo/foo.xml/> > <dc:author>John Doe</dc:author> > </rdf:Description> > </rdf:RDF> > > The xml document that I am pointing at looks like: > > <foo> > <bar1>test1<bar1> > <bar2>test><bar2> > </foo> > > Does this mean that John Doe is also the creator of <bar1> and <bar2>? Maybe. But the level of resolution of your assertion is the XML instance. As pointed out above, you cannot necessarily presume the assertion is valid at a finer resolution. It is similar to saying that, because Joe built a car out of spare parts from the junk yard that Joe also built each of those spare parts. Your statements are about a structure, an XML instance. Whether or not they are valid for the substructure and content of that XML instance is unclear at that level of resolution. > According to, among others, Berners Lee > (http://www.w3.org/DesignIssues/Identity.html), the identifier refers to the > total object described by the node, in RDF terms the 'resource'. Correct. > Consequently, the property does also address every part of that resource. Incorrect. As explained above. > Does this make any sense? Sure. It's a common problem with resolution of knowledge. One of the greatest challenges in getting knowledge-driven systems to interoperate is dealing with differing resolutions. Many times, you have to simply apply poorly founded presumptions and outright guesses, which can be gotten to work for most cases. But even when doing so, one must always keep in mind that usually such presumptions and guesses are dangerous and not to be confused with explicit assertions taken at the intended level of resolution. If you're concerned with proofs, with justifying your decisions based on valid assertions, then you are either stuck with working at the level of resolution of your knowledge, or expressing new knowledge at a finer resolution. Cheers, Patrick -- Patrick Stickler Nokia, Finland patrick.stickler@nokia.com > Kind regards, > Jeroen Bekaert > > Digital Library Research and Prototyping team > Los Alamos National Laboratories > PO Box 1663, MS P362 > Los Alamos, NM, 87545, USA > tel. +1 (505) 664 0580
Received on Tuesday, 8 July 2003 04:37:03 UTC