Re: bNodes and properties (was: Re: evaluating SPARQL w.r.t an RDF query language survey)

On Apr 12, 2005, at 7:08 AM, Seaborne, Andy wrote:

> Dan Connolly wrote:
> ...
>> hmm... why is [] no good? sparqler complains about...
>> select ?R where
>>  { ?R [] [].
>>  }
>> I have only matched things that occur in the subject position.
> ....
>
> bNodes are not currently allowed in the property position beause RDF 
> doesn't allow them there.  Maybe in queries (as variables) they should 
> be.
>
> Turtle does not allow bNodes in the property slot.  Comments?

If Bnodes can appear in the property slot, then the need for the 
distinction between bnodes and query variables disappears. Of course, 
we can still blur them semantically and not blur them 
syntactically...and there seems to be good reasons to keep as close to 
the actual data syntax as possible, including its constraints.

I know conceptually, a sparql triple pattern allows literals in the 
subject position...is that true of the concrete syntax? (I.e., is 
turtle now liberalized that way too?)

I think I tend to come down on the side of folks who want to keep 
aligned with the actual turtle language. I'd like the principle to be 
the minimal set of necessary deviations, obviously marked. As it 
stands, I think dan's fails that test. It adds no new capability; it 
deviates; it's not obvious (to me at least; I always fine []s 
surprising in N3, no matter where they occur).

Cheers,
Bijan.

Received on Tuesday, 12 April 2005 11:53:47 UTC