Comparance of binding approaches: IRI-ID and xml:id/rel.link

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

I tried to compare the different binding approaches.

AFAIK there are basically two approaches on the table:

 I) relative (or absolute URL) in RDF/XML linking to xml:id in the 
content xml
II) The usage of identical IRI in RDF/XML and content



Can we agree on the following statements:

1)
When an ID is given for an element, the ID will identify the ODF element.
The ID should not be changed afterwards, as any change would break 
existing binding.

2)
In case I)  The absolute URL is the location of the ODF element, and can 
be used to locate the ODF element
In case II) The absolute URL is any arbitrary string, without a meaning, 
no way to locate the ODF element

3)
One statement from the blog Bruce gave a link to:
it says it is good to use URLs. Even if the URL might not yet point to 
something existing the user might decide later to create information at 
it's destination.
In case I) the relative IRI would be resolved to an absolute URL, the 
URL always points to the element.
In case II) the created IRI is application not user dependent and will 
not likely be used by an user to resolve information from it

4)
Uniqueness:
I)  Attempt to be global unique by a absolute URL being resolved from 
the relative URL to the xml:id.
II) Attempt to be global unique by a generated URLs for the elements of 
a document.

5)
In case I) RDF subject URL describes one certain ODF element in a 
certain ODF document. The RDF subject changes, when the location of the 
document changes.
[Note: I would see 5) even as a benefit to archive consistency. Nobody 
can guarantee that various copies are still the same and metadata is 
provided for a certain element of a certain document.
A ODF element of a flat document is therefore per se something different 
as the ODF element in a package file, as it is a different entity]
In case II) URI will stay the same, independent of location or package 
form. It is likely that ODF elements will be duplicated and may become 
inconsistent.


Svante