Re: [office-metadata] Question about the IRI of a RDF subject from anODF document

[Michael Brauer <Michael.Brauer@Sun.COM>]

Hi Elias,

Elias Torres wrote:
> Michael.Brauer@Sun.COM wrote on 02/06/2007 11:15:57 AM:
> 
>> The ODF specification already defines hoe reltive URIs are resolved. It
>> says something along the lines that relative URIs are resolved by
>> unzipping the package and resolving them then. This unzip-step is
>> required by all RDF tools anyway, so I don't think there is special
>> processing required.
> 
> Unzip-step is fine. That shouldn't be a problem. However, I'm still not
> sure how does URI resolution work.
> 
> C:\doc.zip
> 
> within [doc.zip]
> 
> images\foo.jpg
> 
> Do you mean resolution to be "C:\images\foo.jpg"? What I mean by resolution

Actually, it would be c:\doc.zip\images\foo.jpg.

But to me more precise: Right now it is only defined how relative URIs 
are resolved. It is not defined how an absolute URI would be 
constructed. But it is straight-forward to extend the URI resolution 
algorithm to include the construction of an absolute URI. That's 
something the main TC could do if requested by the SC.

Furthermore, it is currently not specified how streams in a package 
could be referenced from outside the package. That means, streams in 
other packages than those that contain the URI are not accessible right 
now. But again, that's something the main TC could care about.

> is how to obtain a stable absolute URI that we can use to refer to
> something within the package either in the content or as a file in the
> package. This goes back to Svante's observation from the spec that a triple
> cannot use a relative URI reference.
> 
>> That is to say: If you need a base URI that is different from the URI of
>> the document in RDFa/HTML or RDF/XML+HTML, then we probably need them
>> for ODF, too. If you don't need them in RDFa/HTML or RDF/XML+HTML, then
>> I'm not sure if I understand why you need them for ODF.
> 
> A base URI is definitely needed in many cases by both RDFa and RDF/XML. If
> no xml:base is specified, then the location of the document is used as the
> base. In RDFa for example, the HTML is already found a URL (and hopefully
> that's stable) so relative references work out nicely to other items on
> that page. But for example think of an HTML page that you would like to
> burn to a CD-ROM. Then you save the remote file, add either a <BASE
> href=""> or xml:base and all of the triples in there would be equivalent to
> the remote file.

Sounds reasonable, so we may need a base URI for meta data.

For RDF/XML (or to be more precise, for RDF/XML streams that are 
contained in a package), we get the possibility to specify a base URI 
automatically, since RDF/XML has this feature (that's my understanding 
from what you said about RDF/XML). So there is nothing we need to in 
addition to specifying the algorithm for creating absolute URIs.

For RDFa (or the content.xml, to be more precise), the situation is more 
complex. The ODF schema right now does not allow specifying a base URI. 
We may add this to the schema, but then, the base URI would not only be 
applied to meta data in the content.xml, but to all relative URIs, in 
particular also to whose that reference images. In your above use case, 
that not an issue. But in another thread we discussed the use of URNs. 
If some one sets the base URI to an URN for meta data purpose, then
the consequence is that, all images that are referenced through relative 
URIs become inaccessible. But that's not different than in HTML. How do 
you solve this issue in HTML+RDFa?

Michael
> 

BTW: It is of course another option to define a base URI on the package 
level, for instance in the manifest.xml