RE: CC/PP, RDF and trust issues

At 04:46 PM 5/2/00 +0100, McBride, Brian wrote:
>Graham,
>
>As I understand it, the problem you have stated is:
>
>1) There is an application involving the description of device
>capabilities using RDF.
>
>2) It would be good to include in that description an 'assurance',
>i.e. a statement that some entity is asserting that these are
>the true capabilities of the device.
>
>3) There is a concern about the complexity of reification in RDF
>and you are asking whether there is way to represent this assurance
>without using reification.
>
>4) You are not asking about ensuring the truth of the RDF;
>you believe that is covered elsewhere.

Yes, that's a pretty good summary of where I come from.

>I wonder whether the following would meet these needs:
>
>Consider a resource which is a DeviceCapabilitiesDescription.  It has
>three properties:
>
>         o device - the device being described
>         o capabilities - an anonymous resource which collects the properties
>of the device
>         o assurance - the entity asserting that these are the correct
>properties
>
>This is a simple model and might be represented as an RDF serialization of
>the form:
>
><rdf:RDF
>    xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
>    xmlns:ccppX="...">
>
>    <ccppX:DeviceCapsStatement>
>       <ccppX:device>...</ccppX:device>
>       <ccppX:assurance>Brian Says This</ccppX:assurance>
>       <ccppX:capabilities>
>           <ccppX:colours>2</ccppX:colours>
>           <ccppX:keyboard>12 digit pad</ccppX:keyboard>
>           ...
>           ...
>       </ccppX:capabilities>
>    </ccppX:DeviceCapsStatement>
></rdf:RDF>
>
>I think this is pretty close to your original suggestion.   I don't know of
>anything wrong with it, but I know about as much about knowledge
>representation
>as I do about cryptography (not a lot).

This is roughly my original suggestion, but I see one fundamental 
problem.  It is my understanding that one of the intended goals of RDF is 
anybody can come along and make a new statement about an existing RDF 
resource.  Just like with a natural language, the system is not closed.  (A 
consequence of this is that, in general, there can be no guarantee of 
consistency in a collection of RDF statements:  an RDF subgraph with 
additional constraints is needed for reasoning to be a useful process.)

In the case of your example above, suppose I later wish to add the assertion:

   [ccppX:DeviceCapsStatement]
      --ccppX:capabilities--> [ ] --ccppX:colours--> "3"

This adds a new arc to the RDF graph that appears to be assured as "Brian 
says this", which which (by virtue of the construction sequence described) 
we know is not the case.

(I must say that I wouldn't be able to make these statements based on the 
RDF M&S document alone.  I am combining that with a reading of Tim BL's 
commentary on the intended scope of RDF and the "semantic web" 
(http://www.w3.org/DesignIssues/, if I remember correctly), and comments 
from discussions with other parties.)

Reification, per M&S, works because each RDF statement is assured 
individually, and any new statement that may be added doesn't automatically 
get any existing assurance.  But, as you noticed, I think that explicit 
reification is not sufficiently intuitive to permit easy adoption.

#g

------------
Graham Klyne
(GK@ACM.ORG)

Received on Wednesday, 3 May 2000 04:22:25 UTC