alternative tests - another solution

We have a problem that all agents will pass at most one test from each 
group of alternatives.

We have been discussing modifications to the manifest file to make this 
explicit.

Alternatively we could do the following:

a) provide  in the test doc  a SPARQL query against the manifest.rdf 
files that returns a list of tests that exhibit this. (i.e. the tests 
for which there is another test with the same input doc)

b) provide in the test doc a SPARQL query against an EARL report
merged with the manifest, a list of really failing tests, i.e. the tests 
that have failed for which no other test, with the same input document, 
has passed.

c) (optional) build these queries into the python test harness.

I may try constructing these queries before the telecon

Jeremy