[
Lists Home |
Date Index |
Thread Index
]
Jonathan Borden (jonathan@openhealth.org) wrote:
> > Yeah ok, centralized probably isn't the right word here, and I
> > didn't mean to imply that the language itself is centralized.
> > Just I think there needs to be a way for third parties to
> > associate resources with a namespace in such a way that they can
> > be found using just the namespace.
>
> OK, what are the rules by which this process is carried out. Who gets
> to arbitrate when there are competing visions of what is associated
> with what. Decide this, and you can pick which system to use.
I was thinking like you describe below, priority goes to
resources coming from the namespace in question, but allow for
local override.
> >> On the other hand if you really want to disassociate information about
> >> namespace URIs from mechanisms used to dereference the URI e.g. DNS
> >> HTTP etc, RDF is an excellent choice.
> >
> > Will you talk more about this? The thing that made RDDL
> > attractive is:
> >
> > * it's designed specifically to describe a directory of resources
> > associated with a namespace
> > * nature/purpose are perfect
>
> You say "a" directory of resources? Then RDDL is an excellent choice.
> You can either use the directory provided by the owner of the DNS entry
> associated with the namespace URI or use xmlcat or something similar to
> provide a local override to associate whatever particular directory you
> wish with whatever particular URI you wish.
There are two parts here. The directory of resources is hosted
on any host, regardless of what namespace it is associated with.
Then a server grabs the descriptor and all other descriptors for
a particular namespace and merges them into a complete map of
all resources associated via multiple descriptors. So perhaps
RDDL would work for the distributed descriptors, and use some
kind of simple derivative for the mash.
> The more general RDF idea is that there is an RDF triple store that
> contains a whole mash of triples (associations), not necessarily from a
> single document but perhaps from any number of documents. Now you can
> certainly extact a set of RDF triples from a RDDL document e.g.
> rddl2rdf.xslt in the case of RDDL, if you are looking for the more
> general RDF "anyone can say anything about anything" idea of what
> defines a namespace, then if anyone can add their triples to an RDF
> store, then anyone can define "what is in" the namespace.
Heh, ok, well, I'm almost done with the implementation using
neither RDF or RDDL. I used just vanilla XML, but with a note
that the project at this stage is experimental, and using RDDL
or RDF in the future might be a good option. Using an RDF store
sounds pretty interesting.
Eric
|
