RDF, RDF-Schema and the possible evolutions

Where are we starting from?

• RDF is a data model for the representation of information. It is proposed as the A-box for the web

  • It does not look like this:

    <rdf:RDF xmlns:rdf="..." xmlns:dc="..." xmlns:foo="...:>
      ktrdf:Description rdf:about="http://www.w3.org/">
        <dc:Creator rdf:parseType="Resource">
          <foo:firstName>Tim</foo:firstName>
          <foo:lastName>Berners-Lee</foo:lastName>
          <foo:mailbox rdf:resource="mailto:timbl@w3.org"/>
        </dc:Creator>
      </rdf:Description>
    </rdf:RDF>
              

  • It does look like this

• The W3C RFCore WG is currently active; it's mission includes:

  • define a more precise specification of the RDF data model

  • define a more precise specification of the XML syntax for RDF

  • define a more precise specification of the translation of that syntax into the RDF data model

  • define a precise semantics for the RDF data model.

Where are we trying to get to?

• The star cops scenario
• leaving aside, the voice recognition, natural language understanding, AI planning, .... these are all great and challenging things, but the question that intrigued me is

• What sort of infrastructure do we need to make this possible/easy

• One way:
  • A global naming scheme
  • A common language for representing machine processible information - the A-box
  • A common schema language for describing that data
  • Continuing a trend, from systems, open networks to open information.
  • In effect, we construct a single global distributed database

What questions need answered to get there?

• One hypothesis is that RDF(S) meets needs for constructing this single global database.

Does RDF(S) work and is it Advantageous?

• One hypothesis is that RDF(S) meets needs for constructing this single global database.
• We need to know whether we can build large, highly distributed and decentralised systems using RDF(S)

Engineering - How do we build large, scalable, reliable, performant systems based on RDF(S)?

• Performance is a concern - how do we engineer large scalable systems based on semi-structured data
• How do build in inferencing data stores?
• How do we architect such systems - what are the components? what are the interfaces between them?

How do we program with this stuff?

• Present day RDF tools provide only very primitive operations for programming with this stuff. More powerful primitive operations and more powerful languages for manipulating RDF are required. In this I include RDFPath, RDFQuery etc. What would the APL for RDF(S) look like?

Is it sufficient? Do we need to extend its expressiveness?

• What about DATATYPES?
• For example, there has been much discussion over the summer about representing FOL in RDF - which lacks the basic mechanism of nesting expressions. Do we need to fix that?
• Or do we leave it alone?

Do we have the architecture right?

• Tim Berners-Lee has proposed a layered model for the semantic, not disimilar in form to the ISO 7 layer model for communication stacks
• Is that model correct?
• In particular, if we regard RDFS as lightweight ontology language, have we got the boundary between RDFS and DAML+OIL in the right place?
• What is the relationship to more powerful logic languages

Revisiting Distributed Systems - How do we build distributed systems once we have this capability?

• Object oriented programming has been the dominant programming paradigm for several decades
• Some people seem to believe that the way to build services on the web is to dust off our old ideas about RPC and recast them in XML.
• A semantic web enables to build such systems based on directly manipulating the state of other systems - we should re-examine our assumptions about how to build distributed systems, understand the fundamental nature interfaces between systems and study how they may be recast in the SW

How do we process data we don't trust?

• Conventional data processing systems 'believe' the data they process. This is not something one can do on the web
• How do we handle that?

How do we make this technology accessible to the average web services developer, XML hacker and Web Page designer

• tools
• education
• design principles