Re: cardinality of properties for Assertion

Johannes Koch wrote:

> Well, I asked myself, what an EARL report for OpenSP markup validation 
> would look like:
> A) one Assertion with one testcase property (indicating DTD validation) 
> and many result properties (one for each message), or
> B) one Assertion for each message with one testcase property and one 
> result property.

That's a problem we faced with 'the old EARL'.

The easy way to express it was making each error an assertion.
Indeed, that's one of the main motivators behind discussing
notions of Location at the within-a-page level.

But of course this felt uneasy, not to mention unpleasantly verbose.
So we added a per-page assertion "validation passes|fails".
That gives us two distinct 'levels' of assertion, with the low-level
assertions having only one possible outcome (fails) - because a pass
is an absence-of-assertion.

Now we've finally got a handle on that.  The per-page pass/fail
result is an Assertion, and the individual messages are Evidence
for it in the case of a Fail.

-- 
Nick Kew

Received on Friday, 2 December 2005 10:15:20 UTC