[
Lists Home |
Date Index |
Thread Index
]
> Shelley Powers scripsit:
>
> > Yours is the second interpretation of the model showing v as a resource
> > defined elsewhere. I wonder, though, if a naive person with no
> exposure to
> > RDF/XML would understand to do this? For instance, another
> interpretation
> > could be to nest the second resource directly within the first.
> Would this
> > nesting be illegal? It's perfectly proper XML, but is it proper RPV?
>
> No, it isn't. RPV does not nest (except for the PV element(s)
> being inside
> the R element, obviously).
>
Sorry, John, but I see absolutely nothing in the document that says "thou
shall not nest". Documents must be specific or interpretation comes in and
divergence occurs.
> > According to the Semantics document, there is more to a
> container than bnode
> > and triples. There is an assumed relationship between the
> elements, and a
> > positional constraint.
>
> AFAIK that is just encoded in the container type.
>
> > _2, in order to a) demonstrate that these properties are part of a
> > container, and b) these objects have a positional constraint. Someone
> > walking in off the street without any previous knowledge of the
> RDF wouldn't
> > know do to this, or how to read this correctly just given your
> > interpretation.
>
> Granted. But Tim is serializing triples directly, not working downward
> from RDF/XML.
Sorry, John, but this one really doesn't work. The triples are meaningless
unless we understand how to interpret same. One triple having a bnode is
meaningless to a person reading the XML if they don't know what a 'bnode'
is, and why it's used in this particular triple. There is nothing in triples
that provides meaning -- it's just a serialization format.
This information is addressed directly in the syntax document of the RDF WG,
but not in Tim's document. Now, Tim's document is a just a quick blurb, but
people are already adopting it as a holy grail. If so, better start adding
some specifics now.
>
> > And how would we represent a bnode? Would we show a specialized machine
> > generated code, and if so, how would the person know that it
> wasn't 'real'?
>
> In RPV, an R element represents a bnode iff it does not have an r
> attribute.
> In order for there to be any links to the bnode, of course it has to have
> an id attribute, unlike the situation in RDF/XML where an odd-striped
> element can have neither about nor id attributes.
>
Example?
> > And would the property then be "propertySeq" or "Seq"? This
> would have to
> > formalized, or we'll all be doing something different.
>
> It would be "{rdfnamespace}Seq".
>
Example?
> > For something like reification -- how would a naive user know
> to interpret
> > the reified statement as a set of assertions about a statement
> rather than
> > direct statements? We know, but then, we know the RDF model. This whole
> > thing is based on a naive user being able to read the XML
> without having to
> > know the model.
>
> Not really. RPV is meant to be easier to read in terms of its triples,
> *not* to replace the RDF model with some simpler model.
So what you're saying then, is that a person will have to know the RDF model
in order to read and write Tim's RPV?
I really would like to see some real world RDF models serialized with Tim's
RPV. Has anyone done this, and where can we see them? Links to same?
Personally, I'd love to put these against the RDF/XML version to see -- side
by side -- the differences.
For instance, RSS 1.0 has a container, that would be a good example. FOAF
would be good, too, because it has examples of 'stripes' (node-arc-node-arc
along one path according to the RDF documents, but we can call this
something else if everyone prefers).
Anyone know of an vocabulary using reification?
Shelley
|
