Re: [office-metadata] RDF datatypes

[Elias Torres <eliast@us.ibm.com>]

John Madden <john.madden@mac.com> wrote on 05/10/2007 01:54:37 PM:

> Elias,
>
> The other thing, btw, that I hadn't grasped until recently, is the
> power of the metadata manifest idea. I like many things about it, but
> above all I like the way it allows us to get some of the power of
> Named Graphs by packaging multiple RDF files in a single document
> package.

:) I like it when people recognize the power.

>
> So one related issue apropos inline content is: will the inline
> content all be extracted into one and only one RDF file? In other
> words, will inline markup have the limitation that all assertions
> that are inlined can only be associated with a single Named Graph
> equivalent? I would suspect that this is the case.

Yes. All triples from the content.xml file will go into its own graph.

>
> If that is the case, then it's really important for users of
> documents that might get markup from multiple sources (material from
> multiple metadata authors, marginal comments, speculative comments,
> etc.) to realize that they might tie themselves into a straightjacket
> by choosing to inline all their metadata.

Sure. However, I want you to recognize that this is not a problem with
inline metadata. It's a problem with RDF statements and Named Graphs. It's
the good ole taxonomy problem of which category would you like to assign to
an item. Deciding which NG a triples goes into, it's a big problem. You'll
never have a perfect distribution. The fact is that the content is very
central to your application so you have to create statements no matter
what, and inline avoids content duplication. If everything you ever wanted
to say about a resource could fit and *only* be available in a single NG,
then your world is boring. There will always be more things you want to
refer to that cannot fit into your named graph.

-Elias

>
> I say this as a person who stronglly supports the use of inline
> metadata for some purposes, but also one who is starting to recognize
> its limitations.
>
> John