- From: Graham Moore <gdm@empolis.co.uk>
- Date: Mon, 8 Apr 2002 17:01:37 +0100
- To: "Sandro Hawke" <sandro@w3.org>, "Uche Ogbuji" <uche.ogbuji@fourthought.com>
- Cc: <www-rdf-interest@w3.org>
Two quick comments : the syntax from Sandro is very similar to the XN3 DTD I propose in the paper located on the SWS site. I find this very direct serialisation of the model very refreshing and very powerful. Especially regarding reification - as the Statement has identity and thus its not necessary to re-create the statement in the syntax elsewhere just in order to say something about it. I'm still not sure why a direct serialisation of the RDF model was not chosen as the most obvious serialisation. I mean - the model did come first, right? Which leads me onto the other point that is being discussed further up the thread regarding identity. I would suggest that the identity of a resource is not seperable from the thing that it identifies. Identity is the cornerstone on which we can make assertions about things and between things. If you wanted to have identity as a property, in RDF at least, you would have to assign some identity between the the thing being identified and the identifier. Things are not decomposable into their Identity and the Thing. Related to this seems to be once again the 'am I talking about the web site or the person' issue. Well in my mind we are quite clearly talking about a thing that has identity and nothing more. In the computer system that thing has identity and you make assignments based around that identity. However, you also require a mechanism to bind that identity into the 'real world'. How do we do this, there are a lots of ways; we define ontologies that we agree upon and thus when we bind identities into that ontology we or our software can provide an indication as to what we are talking about. We provide names that people may understand, we make relations that provide context. But identity is just identity. TopicMaps for example provides something directly in the model to disntinguigh the inherent nature of the thing with identity. It provides a way to indicate if an identity is 'just an id' or whether it is the identity of a resource that lives within a computer system. Neither tell me exactly what it is but I have some more information. I'm not sure RDF needs this - its just an example of how one system helps with binding identities to the real world - for real world please read 'the world for which your model is an abstraction'. We should not confuse the assignment of an identity with the posibility that the identity contains information that could lead to some resolution to an abstract entity within a computer system. These two identities serve the same purpose urn:person:gdm urn:ksfskjgjsj:kjsjkfskfks a suitable resolution function MAY resolve this to an entity within a computer system or this may be the identity for just some thing. But both provide a handle by which I can now talk about the thing in my model. cheers gra -----Original Message----- From: Sandro Hawke [mailto:sandro@w3.org] Sent: 07 April 2002 18:34 To: Uche Ogbuji Cc: www-rdf-interest@w3.org Subject: Re: silly question about rdf:about > > Why is rdf:about treated as magic syntax? Wouldn't everything work > > the same in the grammar if > > http://www.w3.org/1999/02/22-rdf-syntax-ns#about were just another > > property? > > I now I'm pursuing a cheap form of argument here, but I can't resist likening > your statement to the old saw: why treat angle brackets as magic syntax? Why > not allow the markup language to define the tag delimiters ... Um, nice try, but languages need magic syntax for separating literal text from structure. RDF/XML doesn't need to give rdf:about any magic status; doing so is just a sometimes-convenient way of thinkings about things. > rdf:about is nothing but a convenience for specifying the subject of multiple > statements in a convenient syntax. It has no standing whatsoever in the > model, or in the concept of the description. That's the way it should stay. Without it (and its twin brother rdf:ID), your "node" remains unnamed, and cannot be merged with another node or otherwise recognized as meaning anything. That is, unless you pick some property as the name property, which is what I'm suggesting. If you reject the primacy of rdf:about, rdf:ID, and essentially-identical other unambiguous naming properties, I don't see how you can use RDF for much of anything. (Since I know you do use RDF for lots of things, I assume there's some misunderstanding here!) > If one doesn't like it, there is always N3. *shrug*. Once one understands how to reify N3 formulas, N3 and RDF/XML are the same, except that you can't serialize some RDF graphs in RDF/XML. (that is, unless we introduce the trick below....) > And actually, it would be nice to get a standard straight triple XML > serialization for RDF. I think Jonathan Borden once posted the obvious > approach. Any reason not to make this official in some way? I can't find Jonathan's, but it seems to me that a clear & simple XML syntax for RDF would turn out to just be an assertion of reified RDF. That is, we define, rdf:Assertion as a subclass of rdf:Statement, and then we can choose to transmit RDF as <rdf:RDF> <rdf:Assertion> <rdf:subject rdf:resource="...some URI-Reference..." /> <rdf:predicate rdf:resource="...some URI-Reference..." /> <rdf:object rdf:resource="...some URI-Reference..." /> </rdf:Assertion> </rdf:RDF> [each of those three properties could have had a literal value instead.] This structure has the advantage of allowing serializations that the traditional RDF/XML syntax prohibits. All it requires beyond RDF M&S is the definition of rdf:Assertion as a subClass of rdf:Statement which has the semantics that any assertion described must be taken to be as true as the description. > If we had such, it would be another way to avoid distraction by serialization > details. I'm with you on that. But some people ARE very focussed on the details of the RDF/XML syntax, and I thought I would nudge them a little more towards simplicity. -- sandro _____________________________________________________________________ This message has been checked for all known viruses by Star Internet delivered through the MessageLabs Virus Scanning Service. For further information visit http://www.star.net.uk/stats.asp or alternatively call Star Internet for details on the Virus Scanning Service. _____________________________________________________________________ This message has been checked for all known viruses by the MessageLabs Virus Scanning Service.
Received on Monday, 8 April 2002 12:02:09 UTC