WOWG: TEST: 2nd draft message

=================

We propose that the test focus area should produce as their first
deliverable to the working group a document that describes one or more test
formats that will be useful for:
- dicussing the properties of OWL
- helping specify the properties of OWL
- testing implementations of OWL

Background:
===========

The RDF Core WG has been using test cases for most of its existence.

The standard RDF test is to assert that a particular RDF/XML file
represents
the same graph as an n-triple file.
Typically both files are small.
e.g.
http://www.w3.org/2000/10/rdf-tests/rdfcore/rdf-containers-syntax-vs-schema/

test002.rdf
corresponds to
http://www.w3.org/2000/10/rdf-tests/rdfcore/rdf-containers-syntax-vs-schema/

test002.nt


Such simple examples have great merit of:
- being clear
- being binary (yes or no)

Thus they are useful for clarifying e-mail discussion.
An example, more or less at random, see Option 2 in this e-mail:

http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0197.html

[[[

<rdf:Description>
  <foo:bar xml:lang="en">foobar</foo:bar>
</rdf:Description>

becomes (please forgive the shorthand):

  _:a <foo:bar>    _:b .
  _:b <xml:lang>   "en" .
  _:b <rdf:value> "foobar" .

]]]

A shorthand syntax is used, we can either agree or disagree with this.
(This example is not endorsed by the RDF Core WG).

This year, a new test format has been being used. This is an entailment
test.

The test consists of a set of files that are the antecedents of an
entailments, and one file that is the consequence.

The test is again binary: does the entailment hold or not.

An example is

( <http://www.agfa.com/w3c/rdf/rdfs-transitive-subSubProperty/test003.nt>
  <http://www.agfa.com/w3c/rdf/rdfs-transitive-subSubProperty/test004.nt>
  <http://www.w3.org/2000/01/rdf-schema#>
  <http://www.w3.org/2001/10/daml+oil#> )
  log:entails
  <http://www.agfa.com/w3c/rdf/rdfs-transitive-subSubProperty/test001.nt> .

The .nt URI's are pointing to NTriple resources which
could/should be "everywhere" on the web (just click
on them to see what they describe).

If we want to have RDFS-entailment, we include
<http://www.w3.org/2000/01/rdf-schema#> which is a
means to refer to the Model Theory RDFS-entailment rules
and similarly for OWL-entailment we include
<http://www.w3.org/2001/10/daml+oil#>

So the log:entails has builtin the appropriate collecting
and dereferencing of uris-to-set-of-ntriples.

The RHS of log:entails is a single uri pointing to a set-of-ntriples
(the ones that should be entailed).

Another example of how such a test is useful in WG discussion is from the
fairly
heated datatyping discussion. The chair summarized many strongly felt
issues
as simple entailments. This reduced the emotional heat.

For example, in a summary message:

http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Feb/0040.html

of the nine or ten disputed issues three are expressed as entailments:
Issue B4, Issue B9 and Issue B10.

So, test cases are not a panacea (six or seven of those issues were not
expressed as test cases), but they are helpful.

Another part of RDF Core's (ongoing) work is the specification of a
manifest
file, that lists all the tests, their type and their status.

At this moment such a manifest file is nothing but "test statements"
(which should be true) as explained in message:

http://lists.w3.org/Archives/Public/www-webont-wg/2002Mar/0022.html


=================

Received on Wednesday, 6 March 2002 15:54:04 UTC