W3C home > Mailing lists > Public > public-rdf-in-xhtml-tf@w3.org > May 2008

Re: 2 RDFa SPARQL Test Harness Issues

From: Manu Sporny <msporny@digitalbazaar.com>
Date: Sun, 18 May 2008 11:58:48 -0400
Message-ID: <48305238.7030109@digitalbazaar.com>
To: Dave Beckett <dajobe@gmail.com>
CC: RDFa mailing list <public-rdf-in-xhtml-tf@w3.org>, Andy Seaborne <andy.seaborne@hp.com>, Benjamin Nowack <bnowack@semsol.com>, Dave Beckett <dave@dajobe.org>

Dave Beckett wrote:
>> The question that we're attempting to ask above is "is there a bnode
>> that has an rdf:type of foaf:Person?", not "is there any node that has
>> an rdf:type of foaf:Person?".
> For that you want.
>  ?a
>    <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
>       <http://xmlns.com/foaf/0.1/Person> .
> }

Thanks, Dave and Andy - we'll change the unit tests to that when we
expect a bnode to be generated.

> I prefer to have "less moving parts" in the testing, just an 
> rdf parser and graph compare.

The decision to use SPARQL was made as I was joining the RDFa TF and I
probably only know a subset of the reasons this approach was chosen. So,
take the following with a grain of salt...

SPARQL is the query language for RDF, and while I can certainly
appreciate the "less moving parts" approach, we wanted to use a language
that was suited for what we were attempting to do: Query an RDF graph
for certain triples and properties.

Much can be learned from chaining these systems together, so while it
does add to the complexity of such a system, the current test suite
reflects how we expect some of these systems to be used in the wild:
feeding remotely generated RDF graphs to SPARQL engine web services. By
doing this, we've been able to uncover certain limitations in our
implementations, such as SPARQL.org not allowing large (4KB+) SPARQL

We had also started out not knowing if we would allow extra triples to
be generated in the [default graph] for conformant parsers. This meant
that we couldn't count on a direct RDF graph comparison, the graph that
was passed in could differ wildly from the minimum conformance
requirements of an RDFa parser. While this is no longer a valid
argument, it did affect the way the test suite was set up initially.

So, those are the two main reasons that I think brought us to where we
are with the test harness. Changing the test harness at this point
doesn't seem like it would buy us much, unless I'm missing something?

-- manu

Manu Sporny
President/CEO - Digital Bazaar, Inc.
blog: DB Launches Medical Record Sales Service with Shepherd Medical
Received on Sunday, 18 May 2008 15:59:30 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:50:28 UTC