Re: (TED] SPARQL, data sources and blackboxes [was (Re: [UCR] ISSUE-12 and ACTION6198)]

Stan Devitt wrote:

> Some examples to consider that don't seem to have come up:
> 
> 1.  What if the RIF compliant application is a custom "editor" designed
> to create or publish a collection RIF rules?  Surely it need not be able
> to execute them.

Yes, this is an important use case that has been mentioned repeatedly, 
but that may not be reflected explicity enough in our use cases. Is 
there something that we should do about that?

What is the impact of that case on the example of a rule containing a 
SPARQL query? Doesn't handling a SPARQL query as a blackbox (that is, 
from the editor point of view, as a piece of free text, I guess) work 
for that case too?

> 2.  What if the RIF compliant application is a broker that farms out
> execution to a collection of rules engines?  Is knowledge that a
> particular engine is capable of executing the collection of rules and
> the ability to pass on the message enough to qualify?

I can imagine this working for dialects: an application on the rule 
retrieving side of the interchange could certainly forward a RIF 
document or ruleset to a specific translator (and thus rule engine) 
based on its dialect (not surprisingly, the application would then 
conform to all the dialects that it recognises and for which it has a 
dialect).

So, why not, for a same dialect, based on some syntactic properties and 
the knowledge that a specific rule language is better to handle them?

But it would still require that the receiving side be able to execute 
the query independently of the publishing side. And, thus, it sets 
requirements on the form of the data available to the receiving side, 
independently of the publishing side, which is what seemed to me too 
strong for RIF Core.

> 3.  What about a A RIF formatter.

What kind of RIF formatter do you have in mind?

Christian

Received on Tuesday, 9 January 2007 13:26:04 UTC