Additional use cases for RDF/A





Since the Best Practices Face to Face Meeting I have been thinking about
potential additional use cases for RDF/A and have, hence, referred back to
www.formsplayer.com/notes/rdf-a.html for current syntax. The following list
is therefore a summary of my current thoughts.

1.    Accessing Sections of RDF in external documents

As discussed with Ralph Swick at the Face to Face, I believe that XHTML can
currently reference external RDF via, <link rel="meta"
type="application/rdf+xml" title="Self-Decription"  href="./example.rfd" />
syntax, for example. This however only provided association for all the
metadata in the linked URI as a single descriptive unit. My feeling is that
the use of ‘external’ metadata should be extended in XHTML to allow:
a.    Explicit extraction of triple collections – RDF, OWL fragments etc,
possibly from multiple sources - for association with, and characterisation
of, explicit XHTML/XML fragments
b.    Dynamic querying of external Metadata sources to allow returned
triples to be associated explicit XHTML/XML fragments
By using the term explicit XHTML fragment I am also, obviously, implying
that an entire XHTML/XML document can be considered as a valid fragment.

This approach would encourage metadata reuse in line with Software
Engineering best practice and the original aspirations for the Semantic Web
www.w3.org/2001/sw/,hence facilitating easy representation of concepts like
… ‘This fragment is one of many in this document that can be characterised
as…’, in a similar manner to concepts currently used in CSS.

2.    Composite IFP’s and Incorrect Derivation of Context via Metadata

I recently had an interesting discussion with Tom Croucher, about just what
the Semantic Web was ultimately supposed to accomplish. During this Tom
passed a rather profound comment, reminding me that IT community as a whole
is somewhat poor at understanding the definition of the word ‘semantic’,
and, given that a significant goal of the Semantic Web is to facilitate the
automated interchange of data, the term ‘Semantic Web’ could be considered
as somewhat misleading within certain technical communities – especially
those who practice traditional data modelling techniques.

Under many circumstances semantic metadata predicates and object pairs
could be better thought of as ‘characterising’, rather than ‘describing’,
their parent subject, as practiced in relational data representation in the
form of name-value pairs. So, where triple collections contain sets of
predicate-object pairs that uniquely identify a subject, their descriptive
capabilities are transformed to provide composite ‘key-like’ qualities. In
making this statement, I appreciate that the concept of semantic
description identification, or keys, is not new and often crops up when
referring to topics like Smushing. Nevertheless my understanding is that
current work in this area only stretches to the concept of an IFP as a
singular predicate-object pair, with <foaf:mailbox> - or its obviscated
sha1 representation - being a good predicate-object IFP example. Having
discussed this matter with DanBri on a number of occasions now, I think
there is a growing consensus that identification through single
predicate-object paired IFP’s is somewhat weak, and that the Semantic Web
will eventually adopt more complex, composite identification schemes in a
similar manner to the relational database world.

So, after all that rambling, what does how does this relate to use cases
for RDF/A? Well, if one accepts that there will be circumstances where
metadata will be used for unique fragment identification within an
XHTML/XML document, and that multiple metadata triples will be used for
this purpose, there are currently no mechanisms to explicitly state which
predicate-object pairs should be used for unique identification purposed
and which should not. Obviously, without such a mechanism, anyone reading
such documents, without prior knowledge of their context, could wrongly
assume key composition and wrongly interpret the fragment being
characterised.

To help alleviate this issue I suggest the addition of an additional
attribute something like ‘metacontext. For example:-

<html xmlns:dc="http://purl.org/dc/elements/1.1/">
    <head>
        <title>On <em>Crime and Punishment</em></title>
    </head>
    <body>
        <blockquote>
            <link rel="dc:source" href="urn:isbn:0140449132"
metacontext=”some_namespace_TBD:identifier”/>
            <meta property="appropriate_namespace:chapter" content="6"
metacontext=”some_namespace_TBD:identifier” />
            <meta property="appropriate_namespace:paragraph" content="32"
metacontext=”some_namespaceTDB:identifier” />
            <meta property="dc:creator" content="Fyodor Dostoevsky" />
            <p>
                Rodion Romanovitch! My dear friend! If you go on in this
way
                you will go mad, I am positive! Drink, pray, if only a few
drops!
            </p>
        </blockquote>
    </body>
</html>

At this point I have to confess that I have never read Crime and
Punishment, so have no idea where, or if, the above quote appears in its
text! Nevertheless the above example is intended to illustrate that the
content bounded by the ‘blockquote’ tag can only be seen as being derived
from paragraph 32 of chapter 6 of ISBN 0140449132, within the intended
context of this document. If this text coincidentally appears elsewhere in
ISBN 0140449132, it is not of relevance. Note also that there is no direct
association between ISBN 0140449132 and the title Crime and Punishment, but
it may be possible to incorrectly infer this, just as I have deliberately
done, from other content within the document.

3.    Metadata about Metadata

By suggesting the inclusion of a ‘metacontext’ attribute this opens up the
whole can of worms around the concept of metadata about metadata…So, I
guess there may well be some merit in promoting metacontext to XHTML tag
status as illustrated?:-

<blockquote>
          <span>
              <link rel="dc:source" href="urn:isbn:0140449132" >
                   <metacontext property=”some_namespace_TBD:identifier”
expires=”10/10/09”
authority=”foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e”/>
                   <metacontext property=”some_namespace_TBD:unbounded”
authority=”foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e”/>
             </link>
                  <meta property="appropriate_namespace:chapter"
content="6">
                   <metacontext property=”some_namespace_TBD:identifier”
authority=”foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e”/>
            </meta>
                  <meta property="appropriate_namespace:paragraph"
content="32">
                   <metacontext property=”some_namespace_TBD:identifier”
authority=”foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e”/>
            </meta>
                  <meta property="dc:creator" content="Fyodor Dostoevsky"
/>
            <metacontext property=”some_namespace_TBD:unbounded”
authority=”foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e”/>
            <metacontext exception=”./metadata_error.html”
authority=”foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e”/ >
         </span>
            <p>
                Rodion Romanovitch! My dear friend! If you go on in this
way
                you will go mad, I am positive! Drink, pray, if only a few
drops!
            </p>
  </blockquote>

The above example is intended to show:
a.    ISBN, chapter and paragraph are intended to act as a composite IFP
for the ‘blockquote’ content
b.    ISBN 0140449132 can only be considered as a valid metadata object
until the 10th October 2009. After that date its use in the composite IFP,
and hence the composite IFP itself, is invalid
c.    It is valid to use (ISBN, chapter, paragraph) and (ISBN) on its own
as attributes for the purposes of smushing
d.    The authority to provide metadata about the ‘blockquote’ content has
been given by foaf:sha1:7eff189e90a0b21cddb1b490ee39524f3cd2ca3e
e.    If any problems occur when processing the composite IFP, i.e.
metadata enclosed within <span>, a default error handling mechanism can be
found at ‘.meterdata_error.html’

Your thoughts on the above ideas would be greatly appreciated.

Kind Regards

Phil Tetlow
Senior Consultant
IBM Business Consulting Services
Mobile. (+44) 7740 923328

Received on Wednesday, 10 November 2004 11:21:44 UTC