Re: aboutEach

Hi Roman,

Thank you for your comments.
[...]

>I've been watching very closely the W3 RDFCore working group. I'm
>developing XML-based knowledge representations for my field field of
>study (programming contests, etc) for a while and found RDF to be a
>good (and standard) way to represent relations that I need.

:)


>I've been almost ready with the moving of all my data to RDF, but was
>set back by the lastest RDFCore resolutions. The problems is: I not
>only want to have a quite flexible representation of my data-graphs
>(entities and relation), but to be able to type and edit (sic!) them
>(mostly) by hand.
>
>So if I have, say 100 resources and I want to attach a common property
>to them (which happens often), then "aboutEach" is the great tool for
>me (actually, "aboutEachPreffix" was even better - you need not waste
>space and time creating bag for them).

A problem with aboutEach was that it was a purely syntactic device, and one 
that the WG felt might confuse the user into thinking it was more than that.

If a syntactic transform is sufficient for your needs, then I suggest you 
consider using XSLT to pre-process your RDF.   If you go down this  route, 
I'd suggest you circulate your solution to rdf-interest and to 
rdf-comments; it would be good if a standard idiom for this sort of 
processing were to emerge.


>More over, it is often the case when I need to have specific
>properties for a couple resources of that 100. Though, RDF spec does
>not say anything about overriding properties (accourding to spec you
>may have the same property multiple times) I was hoping use a simple
>domain-specic rule - always use the last (in document order) property
>value set. Unfortunately, RDF spec does not mandate any particular
>order in which N-Triples are generated from the document, does it?
>(Why not, by the way?)

You are right.  RDF has not concept of document order.  RDF/XML is a 
serialization of an RDF graph.  All the semantics of RDF are defined in 
terms of the graph.  When a graph is serialized, the triples can be written 
in any order, thus document order is not significant to the 'meaning' of 
the RDF.  If order is significant it must be explicitly represented in the 
graph, e.g. by using an rdf:Seq.

To use RDF, you must find a way to represent your information in the rdf 
graph;  I always encourage folks designing RDF solutions to think in terms 
of node and arc diagrams rather than a serialization.


>I undendarstand the reason why "higher level" tools like aboutEach and
>aboutEachPrefix are removed from RDF_Core_, but is there any other
>work going on some "higher level" standard for RDF,

You could take a look at DAML+OIL (http://www.daml.org) which is a 
significant input to the new W3C working group for a web ontology language:

   http://www.w3.org/2001/sw/WebOnt/

Some of the discusion on the RDF rules mailing list

   http://lists.w3.org/Archives/Public/www-rdf-rules/

[...]

I trust this helps.

Brian

Received on Sunday, 3 February 2002 10:33:21 UTC