Re: [office-metadata] Some implementation drafts aside of RDFa

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

Bruce D'Arcus wrote:
>
> We chatted about this off-list, but just to summarize (Elias, feel 
> free to correct me if I'm off and you're there) ...
>
> On Dec 12, 2006, at 11:44 AM, Svante Schubert wrote:
>
>> After a discussion with Michael and Lars, let me bring in some other 
>> implementation details than RDFa.
>>
>> There are various implementations.
>> The earlier used is adding the meta:class attributes for each 
>> element, which has meta data. more than one metadata is attached to 
>> the same element, the attribute value might be a list of space 
>> separated values.
>> Another approach is quite similar to the first, but uses user given 
>> names as attribute names in conjunction with the meta namespace: 
>> 'meta:participant'.
>> And the third is adding to each ODF element which has meta data an 
>> xml:id and bundle these with an indirection similar to the XForms 
>> binding to RDF.
>>
>> <text:list xml:id="meta1">
>>  <text:li xml:id="meta2">Jane Doe</text:li>
>>  <text:li xml:id="meta3">John Smith</text:li>
>> </text:list>
>
> I think you're confusing the issues here. I'd prefer to talk about the 
> model (the RDF triples) and how to bind statements to content nodes.
>
But don't you see, the xml:id is a very simply way to be able to 
identify a content, to which metadata will be related to.
A big advantage against RDFa is that it not require an additional set of 
new attributes to be introduced to ODF only one 'xml:id'. Even more it 
is already a requirement from different side. Can it be simpler, more 
elegant?

> The xml:id attribute provides a generic mechanism to identify a 
> content node as an addressable *resource* (remember, resource and 
> properties), so statements may be attached to it externally.
>
> If we have a table with an xml:id attribute value of "table-1", then 
> it allows you to do:
>
> <rdf:Description rdf:about="content.xml#table-1">
>   ... statements ...
> </rdf:Description>
>
> ... or do to do from outside the package using an absolute URI.
>
> The meta:about attribute is way to attach statements about other 
> resources to a content node (this is where I'm a little unsure of the 
> language).
>
> See:
>
>     <http://www.w3.org/2001/sw/BestPractices/HTML/2005-rdfa-syntax#id0x1dfedc80> 
>
But only if xml:id="table-1" is your desired subject, not if you are 
looking for a set of structures, which in total represent your subject, 
like the example of the user marking a group of elements. Or a couple of 
text passages you mark as important.

>
>> Note: If you want mnemonic xml:id values, there should be certainly 
>> an API to set them, but this is currently out of scope
>
> Correct.
>
>> In the new meta.xml, let's call it xmeta.xml, we have a binding 
>> element similar to the XForms binding, forming the various xml:id's 
>> to a set.
>>
>> <meta:bindings xml:id="binding1">
>>    <meta:binding nodeset="content.xml#xpointer(id('meta1'))"/>
>> </meta:bindings>
>>
>> <meta:bindings xml:id="binding2">
>>    <meta:binding nodeset="content.xml#xpointer(id('meta2'))"/>
>>    <meta:binding nodeset="content.xml#xpointer(id('meta3'))"/>
>> </meta:bindings>
>>
>> Where binding1 and binding2 would be the subject of the RDF.
>
> This seems a needless complication. The subject of a statement is the 
> node as identified by the xml:id or the meta:about attribute.
>
> <text:table xml:id="table-1">...</text:table>
>
> If you want to add statements to it, just add them to the package.
>
> It's really that simple. I don't see the value for the redirection.
The redirection is when bundling a group of elements, like when marking 
content as 'important'.
Or it becomes even more important when the group of elements have itself 
no semantic meaning, only as a group (e.g. a secret message or a future 
ODF feature emulated by metadata of 1.2).

>
>> Question: How is validation being added? If I want to ask if for 
>> example to document being an invitation has the required data and 
>> location?
>
> I see different options if you need to do datatyping.
>
> 1) you add a datatype attribute to the content node directly
Do you like to introduce a new datatype nodeset for ODF and even if we 
had such, what about restriction among given datatypes, like Schematron 
might validate?
Furthermore, schema vary a lot RelaxNG, W3C Schema, etc.. Don't you 
think, that by separating these things from the content, which can be 
changed separately from the content, we become more flexible?
> 2) you include it in the package
Yes or we provide a reference when the information resides outside the 
package
But how do you connect these schema information with content & metadata? 
Why not taking ideas - as the introduced bindings - from XForms, which 
is already W3C standard?
>
> So imagine you have a property called ex:date. Perhaps you have a 
> little configuration file along with your plug-in that gives the 
> information; maybe the actual RDF/OWL schema:
>
> <owl:DatatypeProperty rdf:about="http://ex.net/event#date";>
>   <rdfs:range rdf:resource="http://www.w3.org/2001/XMLSchema#date"/>
>   <rdfs:label xml:lang="en">date</rdfs:label>
> </owl:DatatypeProperty>
>
> You have something else that says the property is required. When the 
> user selects the property there (in their UI they just see the label 
> "date" if it's English), they get presented a little date pop-up to 
> enter it.
>
> That then looks like this in the XML:
>
>     <text:span meta:property="http://ex.net#date"; 
> meta:content="2007-01-10">January 10, 2007</text:span>
>
> I dunno; somthing like that. RDFa here actually helps you validate the 
> datatype of the content precisely because you are binding a statement 
> to specific content.
Looks fine for common datatypes, but what about validating that my 
invitation meta data in the content, which has to have a date and location?
>
> Bruce
Bests,
Svante