Revisiting EARL for Validation

[ Note crosspost ]

I'm making major revisions to Page Valet.

Part of this is a review of the EARL output option.  It seems to me
that the existing use of EARL in validation is something of a mismatch.

Basically, the validator is making exactly one assertion: that a
page is or isn't valid (passes or fails the test of being valid
markup).  Individual error messages are not really meaningful as
EARL assertions: defining the testsubject for them has always been
problematic, and repeating "fails" for each assertion is artificial.

My current thinking is that the validator should make exactly
one EARL assertion, and that the individual messages should be
attached as qualifiers.  They will then fall outside the EARL
vocabulary, but that's IMO preferable to shoehorning it.

So we have something like

<rdf:RDF>

<Assertor rdf:ID="validator">
  bla bla bla
</Assertor>

<Assertion rdf:ID="validation">
<!-- standard EARL-ish stuff something like -->
  <Subject... (the page)>
  <Date ...>
  <TestCase .. (isvalid)>
  <Result ... (pass|fail)
  <assertedBy ... (validator)>
  <note> (whatever) </note>


<!-- and actual errors if any, in their own namespace -->
  <val:error rdf:resource="#err1"/>
   ....
</Assertion>

<rdf:Description rdf:ID="err1">
  <val:line>4</val:line>
  <val:char>52</val:char>
  <val:message>You can't do that here!</val:message>
</rdf:Description>

</rdf:RDF>


Does this make sense?  Any thoughts?

-- 
Nick Kew

Received on Tuesday, 11 March 2003 17:07:43 UTC