Office 1.2 comment: metadata: issues with RDFa

[rjelliffe@allette.com.au]

This comment concerns ODF 1.2 Part 1 s4.2.1 which deals with various xhtml:*

The RDFa specification is referenced, however the RDFa specification does
not actually standardize a vocabulary for use outside XHTML.

This is problematic in four ways.

1) The first is a schema issue. The XHTML specification never defines any
attributes called xhtml:property, xhtml:datatype, xhtml:about and so on.
It defines unqualified attributes property, datatype, about and so on that
may be used on XHTML elements (which have the elements names qualified
with the XHTML namespace.)

This is a long-standing issue that RDF took a convention that is
incompatible with the approach taken by schema languages: the initial RDF
specification made the unwarranted assumption that an attribute name had
the default qualification of the namespace, when used in properties. This
has two problems: first, an attribute may have the same name and different
meanings depending on its context (so its meaning is not univocal as
assumed by the early RDF specifiation). And second, the attributes once
qualified are then specifically non-conforming against the standard
schema: in this case, there is no xhtml:about attribute declaration in
XHTML.

2) The W3C RDFa-in-XHTML recommendation does not provide an adequate
mechanism for use in other standards, unfortunately. The RDFa-in-XHTML
scope statement says "This document specifies how to use RDFa with XHTML."

In section 1 of the RDFa-in-XHTML specification, which is only
informative, it says: "This specification deals specifically with the use
of RDFa in XHTML, and defines an RDF mapping for a number of XHTML
attributes, but RDFa can be easily imported into other XML-based markup
languages."

However, no such standard mapping exists. This is made explicit in the
RDFa Overview (currently a draft)* which states in S1: "RDFa itself is
intended to be a technique that allows for adding metadata to any (XML)
markup document, including SMIL, RSS, SVG, MathML, etc. Note, however,
that in the current state, RDFa is being defined only for the (X)HTML
family of languages."

I had to investigate this issue last year for a government project our
company was doing. We had intended to add RDFa to an XML format. I
conducted an email exchange with various parties at W3C mailing lists, and
they were very helpful. My conclusion from things like this exchange** was
that very basic syntactical issues had not been sorted out: RDFa is not
baked enough to be useful for XML.

W3C's Ralph Swick confirmed that defining RDFa over non-XHTML languages
was not a capability that had been addressed***  They just had a general
"intent".

3) The third issue is not really technical. It is well-known that the
HTML5 specification is highly influenced by technocrats who are
(technically) antagonistic to XHTML, and to its RDF metadata in
particular. The editor's version of HTML5 from the WHATWG site for example
does not have the RDF attributes. The editor Ian Hixie is very keen on
microformats instead.

It is quite possible that RDFa will not be a technical winner even in the
browser space. I do not wish to get involved in sides (I think RDFa
*would* make sense for XML but also I think microformats make more sense
for plain HTML too), however, I want to flag that not only is the current
W3c RDFa recommendaton technically incomplete  for use outside XHTML, it
is also the subject of controversy within HTML.

4) In any case, the attributes in the xhtml: namespace are not actually
defined by XHTML. So even if it is acceptable that the RDF sloppiness in
treating unqualified attributes as if they were qualified by their
namespace, the RDFa attributes are really an extra unqualified vocabulary
in no namespace, which have special significance to an RDFa-in-XHTML
processor. So it is still inappropriate to treat them as if the attributes
were defined in XHTML let alone qualified by its namespace.

SUGGESTION

What can be done? I suggest there are three possibilities given as A,B,C,D:

A) Include a note that despite the use of the XHTML namespace, the
attributes do not in fact belong to this namespace. Therefore validators
must not look to the W3C schemas. I don't like this because the purpose of
namespaces is to *prevent* third parties redefining elements or
attributes. It is bad citizenry to perpetuate the initial RDF mistake in
this regard.

B) Remove the xhtml:attributes. Treat them as extension attributes, not
available in strict ODF. Engage with the W3C RDFa effort to get a proper
spec out, then adopt that, and fold it into ODF at some future version.

C) Repeat the approach taken with SVG, where the fact that a close dialect
is being used was made explicit by using customary prefix, but an ODF
namespace URI. I think that is the correct approach in that situation. I
would suggest the prefix "rdfa" would be better to use, with the
appropriate text.

RDF's Dave Beckett commended GRDDL to me in this kind of situation****: it
is a very simple transform that generates 'real' RDF out.  I suggest that
the ODF specification should include a normative GRDDL transformation to
RDF triples (Not RDF XML).

D) If the decision is made to adopt GRDDL, then there is no particular
reason to stick with RDFa concepts. The ODF TC can remove RDFa
complication from the text, in particular any confusing references to
xhtml, do whatever it wants with the syntax, and provide the appropriate
GRDDL mapping. The notational RDFa references can be added at some future
time after RDFa-in-XML recommendations are made at W3C.


In any case, a few extra paragraphs are certainly necessary to justify the
approach taken. The prudence of avoiding adopting third party standards
before they exist or are debugged should be self-evident.

Cheers
Rick Jelliffe


* http://www.w3.org/2006/07/SWD/RDFa/rdfa-overview
** http://www.mail-archive.com/public-semweb-lifesci@w3.org/msg05219.html
***http://lists.w3.org/Archives/Public/public-semweb-lifesci/2009Jun/0091.html
****
http://lists.w3.org/Archives/Public/public-semweb-lifesci/2009Jun/0080.html