Re: RDFa with multiple CURIEs as property..

Hi Mark,

Mark Birbeck wrote:
> Hi Svante,
>
>   
>> No longer in need, Dan Brickley was so kind to show me the feature has
>> merit.
>> Since often rdf vocabularies overlap in scope, and instances might have
>> multiple audiences.
>>
>> Therefore I will suggest the ODF TC to overtake the property.
>>     
>
> That's great.
>
> But I would reiterate that I believe you should adopt RDFa in its
> entirety, rather than looking for opportunities to fork the
> specification, leading to the possibility of confusing people.
>   
> In other words, whilst it's of course good news that you now see the
> merits of some particular feature, that should not be relevant when it
> comes to the question of language integration.
>
> Sorry to sound hard-nosed about this, but I think it would be really
> counter-productive if we were to go through exactly the same loop
> again next week, when some other feature doesn't seem obviously
> useful.
>
> ...just take the whole shebang! :)
We had very long emotional discussions about this topic in the ODF 
Metadata Subcommittee.
The final decision was made involving Elias Torres a strong promoter of 
RDFa.
I am not able to reopen the discussion, but let me try to explain you my 
result from my personal view.

We (Odf applications) are facing a different environment first the 
applications are usually not DOM applications, the XML is just a 
serialization of their arbitrary model (in OpenOffice.org C++ classes).
Second our file format is a ZIP that is able to bundle arbitrary RDF files.

By having the ZIP we have the opportunity to follow a design principle 
to separating the RDF graphs and content under control of the ODF 
application as much as possible would ease the processing for all.

We decided to use an RDF metadata manifest file (some kind of an index 
for all metadata in the package) to describe the metadata files in the 
package (to have one defined spot to look into).
Under control of the ODF application Metadata plugins can query if there 
is interesting metadata available for them.
(BTW we bundling Redland with OOo 3.1).

To describe a certain ODF element within an RDF statement, we give the 
element an xml:id and provide in the metadata manifest a mapping to an 
IRI for that element with the xml:id.
By this we had the mapping of the two worlds XML and RDF.

Nevertheless we were still very much in need of RDFa as instead of 
duplicating the literals of the content in the rdf graph and risking 
inconsistencies, we used RDFa to embrace the content into the graph.
A wonderful technique we were glad to overtake.

This is our first approach to use RDF in the office, let's see how it 
works out and we still can expand it in the future.
Let's do baby steps..

Currently we are using only those attributes, that can not be expressed 
within an RDF file.

xhtml:about : The IRI of the RDF subject
xhtml:property : The IRI of the RDF predicate
xhtml:content: The RDF object, if it appears. Otherwise, the literal 
text content of the OpenDocument element is the RDF object
xhtml:datatype: Data type of xhtml:content

Hopefully I make you our decision clearer, we are definitely not 
forking, we are just using a subset.
If you think about modularizing the spec, this can be a candidate, let's 
call this a module ODF is implementing.

Thanks to all of your group for the great work you have done to provide 
us RDFa!
And just in time you become a Recommendation so ODF 1.2 is finally able 
to reuse your namespaces and CURIE mechanism that will save us a lot of 
space in large spreadsheets..

An happy Halloween, Mark!
Svante

Received on Friday, 31 October 2008 19:31:50 UTC