Re: [office-metadata] Reuse of metadata proposal for non ODFapplications

[Svante Schubert <Svante.Schubert@Sun.COM>]

Bruce D'Arcus wrote:
>
> On Jun 24, 2007, at 12:15 PM, Svante Schubert wrote:
>
>> There has been earlier some discussion and tendencies about making 
>> our metadata proposal for packages more modular, more reusable for 
>> other non ODF applications.
>> As there were no opinions against this approach, we should come 
>> quickly to a proposal how this can be established.
>
> Well, but I think we need to the TC to say they want this before we do 
> anything about it.
Sure. Thought we could cascade from the SC to the TC. If the SC declines 
already, we do not have to bother the TC.
>
>> Suggested changes in detail:
>>
>> 1) In content metadata namespace change:
>> http://docs.oasis-open.org/opendocument/meta#
>> to
>> http://docs.oasis-open.org/package/meta#
>
> I get a 404 on the root package URI, so we'd need to address that. Who 
> would control that URI? The ODF TC?
I assume OASIS would control the URLs above. They should be created as 
soon our proposal has been approved.
Although it might be wise to ask back earlier at OASIS on this topic, 
right after the TC agrees on the idea of this change.
>
>> 2) Metadata manifest namespace change:
>> Changing the namespace of basically the complete odf: prefixed RDF 
>> vocabulary from
>> http://docs.oasis-open.org/opendocument/meta/package#
>> to
>> http://docs.oasis-open.org/package/meta/manifest#
>
> ??? Why the "manifest" at the end? Isn't that redundant?
It is the RDF vocabulary used by the metadata manifest.
>
>> But still remaining the odf: namespace, which would be slightly 
>> changed accordingly from
>> http://docs.oasis-open.org/opendocument/meta/package#
>> to
>> http://docs.oasis-open.org/opendocument/meta/manifest#
>>
>>
>> With the exception of the ODF related elements, which are:
>>
>> odf:ContentFile - the OpenDocument content.xml
>> odf:StylesFile - the OpenDocument styles.xml
>> odf:Element - an OpenDocument XML element
>>
>> We should introduce for the package an
>> xml:Element for XML elements, from which odf:Element is a subclass.
>> As an xml element has no namespace per se, I would suggest 
>> "http://www.w3.org/TR/REC-xml#Element";
>
> Hmm ... I have a feeling this might be a little dicey. Not sure you 
> can really do this.
If we do something in a consistent manner we can do a lot.
But sure we might ask back at the W3C. Any better idea to classify an 
XML element in RDF is certainly welcome.
>
>> Finally if it is anyhow conformable with RDF, I would suggest a 
>> generic approach to identify each kind of ODF element by an rdf:type 
>> similar to its IRI given by namespace and local name.
>> For consistency reasons of RDF, I would add two minor adoptions 
>> before reusing the IRI. First insert the delimiter '#' between 
>> namespace and localname if not existent and second change the first 
>> letter of the local name according to RDF classes to a capital letter.
>
> I don't know, this feels problematic too. We're going to reuse 
> existing IRIs, except not really? Those existing IRIs are not resolvable?
>
>> For example, some table element in the document could be identified 
>> in the metadata manifest according from the IRI resulting from 
>> table:table.
>> The RDF class for table:table would be 
>> "urn:oasis:names:tc:opendocument:xmlns:table:1.0#Table" after the 
>> adoptions instead of the original 
>> "urn:oasis:names:tc:opendocument:xmlns:table:1.0table".
>>
>> <odf:Element rdf:about="uri:someIRI" idref="someID">
>>    <rdf:type 
>> rdf:resource="urn:oasis:names:tc:opendocument:xmlns:table:1.0#Table"/>
>> </odf:Element>
>>
>> Could be written shortly as
>> <table:Table rdf:about="uri:someIRI" idref="someID" />
>>
>> The subclassing of odf:Element from all existing ODF namespaces 
>> declaring OpenDocument elements would be done in the ODF related draft.
>
> I'm curious if Elias finds the time to comment on this, but this just 
> seems wrong to me. It's basically a hack to create more human readable 
> and namespace-prefixable IRIs by getting around the fact that the 
> original namespace IRIs for ODF 1.1 weren't designed with RDF in mind.
Do I understand it right, that you like the idea of adapting RDF classes 
from element namespaces and local name, but you dislike the modification?
After thinking it over, I think you are possibly right about dropping 
the modification.
Applications might in general get the local name from the IRI by 
removing the namespace found in the by odf:path/pkg:path referenced file.

Svante