Re: Evidence Class

On Mon, 19 Sep 2005 19:10:14 +0200, Shadi Abou-Zahra <shadi@w3.org> wrote:

>
> Hi,
>
> During the F2F back in March [1], Charles had suggested [2] the  
> following class for evidence:

Well, it is a structure, not specifically a class...

> <earl:evidence parsetype="Collection">
>   <earl:ruleSet rdf:resource="ShadisOWL"/>
>   <earl:Assertion rdf:resource="someAssertion"/>
>   <earl:Assertion rdf:resource="anotherAssertion"/>
> </earl:evidence>
...
> The earl:ruleSet was introduced to allow pointing to a rule set. For  
> example, a set of OWL constraints that have to be met for the evidence  
> clause to be true. However, I think we will be opening ourselves to a  
> whole lot of complications with this.

Can you elaborate on what complications you think we will be opening  
ourselves to? Obviously there is no single agreed rules language for RDF,  
so interoperability depends on being able to understand the things pointed  
to. But in many cases the rules are going to be simple OWL, I believe -  
simple enough that we could describe it in the spec.

> On tomorrow's teleconference I'd like to go over the following:
> * Refining the class and discussing the earl:ruleSet property
> * Agree on where to put it (directly in an assertion vs result)
> * Approve it to be added to EARL 1.0 Schema in the next update

The minutes of the call state " ruleSet is a can of worms which could be  
scary for us to introduce " yet provide zero evidence to support the  
statement. If this is something more than mere FUD, please elaborate,  
since it appears that the meeting almost decided (there is no decision  
recorded, so I can only assue that one has not been taken yet) to remove a  
property without testing the use case based on this single unsupported  
(except by repetition) statement.

The evidence construct does not remove the need for test mode. It applies  
in cases where the test mode is by inference (whose URI happens to end in  
heuristic - proof that relying on URI strings to be meaningful is a dumb  
idea ;-), but would point to tests done manually or automatically.

Indeed, I would expect some of the rules used (basic OWL constructs,  
mostly reusing things we already have in the current spec draft) to  
actually discuss test mode - for example a restriction on WCAG 1.1 results  
that only trusts manual verification for a pass, but is prepared to accept  
automatic verification for a fail, seems an obvious thing to do.

I think it makes sense to make it a property of the Assertion, since that  
is where the test mode property goes...

cheers

Chaals

-- 
Charles McCathieNevile                      chaals@opera.com
          hablo español - je parle français - jeg lærer norsk
         Web dreams are free:   http://www.opera.com/download

Received on Saturday, 24 September 2005 13:27:28 UTC