Re: expected result -- too simple?

On 10 Sep 2007, at 02:22 , Michael Kay wrote:

>
> If your goal is "to see whether different processors construct the  
> same
> schema from the set of schema documents, or different schemas."  
> then perhaps
> the facility you're missing isn't the ability to make assertions  
> about the
> PSVI, it's rather the ability to make assertions about the schema
> components; and rather than having an XML representation of the PSVI,
> perhaps we need an XML representation of the schema components.

Yes.  The last time we looked at that, we were unable
to reach consensus on it (or even on whether such a thing would
be desirable), so it has been a kind of unsolved
problem.  But in the context of the test suite, we don't need
a standard way for validators to expose the schema components,
only a standard way for the test suite catalog to describe
them.  And enough time has passed that maybe it's worth looking
at the question fresh.

> Either way, you seem to be asking for a substantial increase in the
> complexity of the test framework, which is not necessarily a good  
> thing, at
> a time when we seem be stretched for resources; additional  
> complexity in
> some of the tests would seem preferable. Presumably you can  
> automate the
> construction of your 81 tests.

I can certainly do that, and will.

I do think that in the long run, the test suite would be more
useful if it specified the expected PSVI for each instance
in full, rather than just the [validity] property on the
validation root.  That, in turn, would require an XML
representation of the PSVI, and everything said above about
earlier attempts and lack of consensus applies to that, as
well.

Michael

Received on Monday, 10 September 2007 15:36:34 UTC