Re: [office-metadata] Our discussion on the Wiki example

[Bruce D'Arcus <bruce.darcus@OpenDocument.us>]

On Dec 4, 2006, at 12:46 PM, Svante Schubert wrote:

> But what really bothers me is that it was designed for XHTML being a 
> flat format. RDFa is about embedding the meta data. ODF is compound 
> and anybody correct me if I am wrong, I was sure everybody figured out 
> that these should be separated (without redundancy).

This is indeed the difference.

But note that I think -- for all the practical reasons that Elias and I 
demonstrated -- we can't be puritanical about the no redundancy idea. 
At minimum, we need to be able to include presentation content in 
content.xml.

For the citation example, one worry I have about moving all the logic 
into the metadata file is the copy-and-paste problem.

> We might as well step back and define the scenarios, we would like to 
> show in Wiki as examples:
>  For instance:
> 	1 	metadata contains/reference additional data
> 	2 	metadata specifies a unique content
> 	3 	metadata specifies a class of content
>  and collect basic design decisions we agree on, like
> 	1 	No redundancy (no repetition of data from the content in the meta 
> data)
> 	2 	RDF compatible
> 	3 	generic solution / coverage of use cases
> 	4 	simple solution
>  And afterwards we make proposals perhaps based on implementation like 
> RDFa.
>  Analyzing their dis/advantages and choose one. Does not sound to 
> complicated nor time consuming.
>  As Bruce was so kind to start with one example, I commented it asked 
> for changes. I see no delay with this process.

It depends. Certainly I can see three options for the citation case.

1. all field logic moves to the metadata file (as the example we 
discussed)
2. the field logic stays in the content.xml file in a new field, but 
using RDFa to encode it
3. as above, but using using specific XML elements for the encoding

Generically, I see two options:

a. an all RDF/XML approach
b. a hybrid RDFa and RDF/XML approach

The a option would still rely on being able to attach URIs (local or 
global) to content.

If we get beyond that, and start considering a number of other options, 
then we'll easily take six months to get a proposal.

Remember: it took us 9 months to collect the use cases and derive the 
requirements. We now have two more to deliver the draft proposal. That 
is not a lot of time.

...

>> <table name="table1">
>> ...
>> </table>
>> </body>
>>
>> Also, is ODF content/source copy and paste a requirement for our 
>> metadata
>> proposal? I didn't think it was. I hope we are not expecting people 
>> to hand
>> write ODF (e.g. no need for mnemonics).
>>
>
>  Mnemonic approach is helpful for the writer, should be recommended, 
> but is and can not requested.

"Helpful" for what "writer"? A user will never see these, and a machine 
doesn't care. What is most important is that the objects be uniquely 
identified such that they can be reliably referenced. Tha's all that 
really matters.

>  I prefer - as already stated - the approach of attribute references 
> between content.xml and metadata, which is not one of your approaches 
> above - not #1 nor #2.

Well, the question is the direction. His second example above simple 
gives the table a name ("table1"), and one references it in the 
metadata file ("content.xml#table1").

>  In content.xml:
>  ============
>  ..
>  <text:p meta:class="date">
>      <text:span meta:class="month">May</text:span>
>      <text:span meta:class="day">8th</text:span> at
>      <text:span meta:class="time">10am</text:span>
>  </text:p>
>  ..
>  [NOTE:
>  I changed meta:id to meta:class to avoid the impression, that meta:id 
> is unique. The naming 'meta:class' is not important for now.
>  And the value of meta:id is just an arbitrary string. But here only 
> provided as mnemonic default string by a brave plugin programmer. ]
>
>
>  In meta package:
>  ============
>  something RDF compatible
>
>  This is a very simple approach. Everything seems to be accomplished 
> by it, what advantage have #1 or #2?

The advantage of this approach (metadata --> content) is that it's also 
the same for Rob's "extrinisic metadata" use case. It also involves no 
change to the content, since tables already get an id (the table:name 
attribute) that is local to the document.

And the model is clear:

	<content.xml#table1> a odf:Table .
	<content.xml#table1> dc:title "Some Title" .

I'm not clear what you're modeling in your example Svante.

...

>  Drafted in two sentence in general is my hopeful wish in linking the 
> following:
>  I would like to be able to create a link to a document pointing to a 
> certain semantic not to a structure.
>  Like pointing to the node set of all XML nodes having a certain class 
> of meta data like Bruce's citation.

Certainly with RDF you can define any property you want; e.g.:

	<ex:xpathForAllCitations>//cite:citation</ex:xpathForAllCitations>

Likewise, you can have a property to represent an xpointer.

That doesn't mean we create some specific structure just to do that; 
does it?

We have a focused set of requirements that the TC happily agreed to. I 
hope you aren't proposing to add a new one?

Bruce