Bug 005: Reification Trick

This is feedback on a Last Call Working Draft:

Evaluation and Report Language (EARL) 1.0 Schema
W3C Working Draft 10 May 2011
http://www.w3.org/TR/2011/WD-EARL10-Schema-20110510/

It is natural, in the RDF model, to consider that in the case of a
test, some system being tested will have some level of conformance
against some test criterion. In other words, some system will pass or
fail a test. In the EARL model, other verbs are possible, but note
that they are verbs:

Firefox 3.5 passes CSS test case 1.026.
Apache fails HTTP request test case 730.

In EARL, however, these verbs are used in the object position! In
other words, an Assertion consists of a TestSubject (the subject), a
TestCriterion (the verb), and a TestResult (the object), and the
result is what includes the pass or fail information. Obviously there
are other result states which I omit for clarity.

In fact it is not so clear that there is a subject, verb, and object
relationship here, and this is what this bug is about. In fact we
started out in earl by using a reification trick. We said that an
earl:Assertion was a sub class of rdf:Statement, which is a triple.
Then earl:subject, earl:test, and earl:result (in the current schema)
were sub properties of rdf:subject, rdf:predicate, and rdf:object
respectively.

This resulted in a lot of debate as to the approach,* because of some
problem with the difference between RDF's reification method and the
one used in the SWAP/N3 environment, which has true contexts, also
called formulae in the code.

I do not know what the state of the art is in this debate. But I am
surprised that the simplicity of the design, in a test passing or
failing some criterion, has vanished. I think it would be much clearer
if it were brought back. The existing approach should at least be
rationalised, if there is rationale behind it; the rationale must be
stronger than for my position.

Even if this reification trick is not reflected in the schema, say, it
can be reflected in explaining the design of the language. It would
help the Guide out too, but it would be primarily helpful in the
Schema.

If you like, a summary of this issue could be that ยง 2.5. TestResult
Class should be considered predicative to some unspecified but
manifest degree, whereas it currently is not.

* Later on. But originally we had to ask the RDF Logic group over and
over again for feedback on whether it was a valid approach. I bugged
W3T people, but I was mystified over it for ages. I think the problem
was that the number of people who had the expertise to solve this was
small, and their attention was mainly elsewhere. There may still be
not that many people who have the relevant expertise here.

-- 
Sean B. Palmer, http://inamidst.com/sbp/

Received on Tuesday, 10 May 2011 22:25:18 UTC