Re: ODF and semantic web

Hi Mark,

On 10/14/07, Mark Birbeck <mark.birbeck@formsplayer.com> wrote:

...

> I note that the attributes used in ODF are 'inspired' by RDFa [1]--but
> why not just incorporate RDFa as is?
>
> It's especailly confusing for authors when this 'inpiration' seems to
> involve copying some RDFa attributes, but changing the names of
> others. For example, @about is used, but @datatype has been renamed to
> @data-type!

The explanation of why we didn't use RDFa as is goes back to what I'd
call the unique challenges of integrating RDF into an already deployed
XML format for productivity applications.

A few issues off the top of my head :

* all nodes in ODF must be namespaced; the RDFa attributes are not

* uneasiness about CURRIEs in attributes (we require the full URI)

* the big one: using the short-hand syntax in the context of an
editing application. ODF is rarely going to be hand-authored, and
implementors thought it too large a burden to be responsible for
tracking the subject URI apart from the rest of the triple in the
context of a dynamic content environment, where users are going all
kinds of funky things. So, for example, we require the subject and
predicate URI to be on the same node.

I don't recall the datatype vs. data-type issue, but it is probably a
consequence of consistency with existing ODF patterns. I can check on
that, as the details are not set in stone.

> This lack of alignment is a shame, especially when the proponents of
> ODF are generally critical of the confusion that can be caused by
> companies and organisations pursuing alternate document formats. There
> is a fantastic opportunity here for creating tools and search engines
> that could leverage a 'standard' way of incorporating metadata into
> HTML, XHTML, ODF, and other mark-up languages. That opportunity now
> looks like it is going to be missed.

That's too strong, and is only the case if you believe RDFa is the
only way to integrate metadata into content.

The in-content ODF metadata syntax is a subset of RDFa; one can easily
write an XSLT that maps it perfectly to RDFa (or RDF/XML, TriX, etc.).

I think the expectation is that the most common case will have the
triples serialized as RDF/XML in the file package.  For example, my
focus is citations. In that case, we'll have a simple field with the
formatted text, but the citation metadata will get stored as RDF/XML.
Once one gets into more complicated metadata scenarios, this would
seem the preferable approach.

We've adopted a kind of minimal syntax for in content statements where
the object is a literal. I can imagine in the future this subset might
be expanded as implementors gain experience.

So I guess recognizing the above, are there any changes you think are
important for us to consider now?

Bruce

Received on Sunday, 14 October 2007 12:33:29 UTC