Re: [Test] Review (ACTION-661)

Gary,

Thank you for your review.

public-rif-wg-request@w3.org wrote on 12/01/2008 05:56:44 PM:

> 
> some comments on the proposed FPWD of the RIF Test Specification:
> 
> 1. We need an XML schema instead of or in addition to the RDF schema. 
> Many implementors will not understand the model theory and will turn to 
> the test cases ... and again be stymied when they hit this schema.

Sandro brought up a similar point. It sounds reasonable to me (the RDF 
format of manifest files does not need to be very complicated, but may be 
unfamiliar to the target audience).

> 
> 2. Why do we say these are BLD tests when many are actually Core tests?

The current version of the document doesn't address Core because Core 
wasn't finalized until the same time Test was to be frozen. We could do it 
now (update the document, repository, and (approved) test cases) by 
Thursday evening, I think, and then those updates should be reviewed. 

Whenever these updates are done - once we label test cases as Core, I 
think we also need to say they are PRD (as well as BLD) test cases? For 
the purpose of test cases, I was thinking to treat safe-Core as a dialect.

> 
> 3. We have made the task of creating a test harness difficult.  The 
> conclusion and non-conclusion "documents" are not legal RIF documents. 

The decision to represent the (RIF) conclusions as RIF condition formulas 
was made because BLD defines conformance in terms of entailment of 
condition formulas and not document formulas. 

We have some residual references to conclusion "documents" in the test 
document and we should reword these if we keep conclusions as condition 
formulas.

Some of the test case conclusions (e.g. 
RDF_Combination_Constant_Equivalence_Graph_Entailment_1) are not RIF, they 
are RDF.

> I would expect many RIF implementations will have either a query 
> facility or a print builtin.  (unfortunate that we specify neither.)  I 
> would expect to use either a query or a print builtin to implement the 
> test harness.  I would combine the conclusion "document" wrapped with a 
> query or wrapped with a rule that prints "pass" (or "fail" for a 
> non-conclusion).  But then, that makes  the conclusion part of the 
> premise document, and thus tests such as Local_Constant will fail.  I 

If the method of combining is to import a constructed document, then I 
think the Local_Constant test case will be ok, because the local constant 
in the imported document will be distinct from the one in the premises 
document.

> think we should give some guidance that these would be examples of how 
> to implement a test harness (provided it can be made to work), and we 

We can add something about this in section 6.

> should explain (if we can) why these are not part of the standard.

Is the test document the right place to explain why print and query aren't 
part of the standard?

Stella