Re: EARL Semantics and Queryability [was: Re: EARL-producing testing tool]

> [...] I assumed that the association of a context (date,
> author, etc) with a given result (pass/fail/...) would hold
> when the graphs were merged.

Well, some bits hold and some bits don't - and it also depends upon
the type of query. For example, you can reduce a bit of EARL so that
it loses all of the context info using a filter/rule. In my previous
email, I reduced the statements from the following evaluations:-

[ earl:asserts [
   rdf:subject [ rdf:type earl:WebContent;
     earl:reprOf <http://www.w3.org/blargh>;
     earl:date "2001-10-15" ];
   rdf:predicate earl:fails;
   rdf:object :MyTest ] ] .

[ earl:asserts [
   rdf:subject [ rdf:type earl:WebContent;
     earl:reprOf <http://www.w3.org/blargh>;
     earl:date "2001-10-17" ];
   rdf:predicate earl:passes; # typo in original
   rdf:object :MyTest ] ] .

RDF Core seem to be in the process of deciding that one reified
statement is not necessarily the same as another, even with the same
subject, predicate, and object. That's consistent with the way in
which I reduced the statements - it's intutive, and it works well with
EARL.

Of course, we don't actually reify all of the statements in the model.
We only reify the actual earl:Assertion itself, because it's all that
we need to reify; the others are objective truths, whereas an
earl:Assertion is subjective, and varies from tool to tool. This
method has been up in the air for some time now, but no one seems to
have challenged it, modulo DanBri's hypertext-in-RDF idea.

I'd certainly like some discussion of the model - which I've been
pondering for some time now - by the group, although everybody seemed
happy with it from day one. In fact [credit where credit's due], the
model came after an entire weekend of work, and hours of discussions
with Aaron Swartz - in which he proposed [1] the model that we had in
EARL 0.9. The model in EARL 0.95 is basically the same, except that
the predicate and object in the assertion are swapped around since it
seems to make a little more sense that way.

I can recall discussing the "only reify the Assertion" idea before
somewhere, but it's pretty difficult to find specific discussions on
the list. It's a bit sad that 90% of the design rationale for EARL is
in the lists... ugh. Then again, at least it's all archived.

[1] http://lists.w3.org/Archives/Public/w3c-wai-er-ig/2001Apr/0033

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://purl.org/net/swn#> .
:Sean :homepage <http://purl.org/net/sbp/> .

Received on Monday, 18 February 2002 22:56:50 UTC