- From: Andy Seaborne <andy.seaborne@epimorphics.com>
- Date: Thu, 29 Sep 2011 12:02:48 +0100
- To: public-rdf-dawg@w3.org
On 28/09/11 23:01, Gregory Williams wrote: > I'd like to make three changes to test suite files, and wanted to run > them by the group. > > 1. > > In test csv-tsv-res/manifest#csv03 I'd like to change the lexical > form used in the CSV results file (csvtsv03.csv) and the source data > file (data2.ttl) for the xsd:double value to use canonical form > (changing from "1.0e6" to "1.0E6"). While this doesn't totally solve > the problem of the csv serialization being lossy and making > value-based comparisons difficult, if we're going to choose to use > either canonical or non-canonical form, I think canonical form is the > better choice to allow canonicalizing implementations to more easily > pass the test. Fine by me. > > 2. > > I'd like to remove test service/manifest#service5 from the manifest > list as it tests the "SERVICE ?var" form that I believe we've > resolved not to specify. An alternative would be to mark this test > with an mf:requires and mint a new IRI to represent the > variable-endpoint extension. The proposal to make (service) a binary operator woudl address teh key issue. > > 3. > > I'd like to change the SERVICE manifest to mark federation tests as > using the service "feature", not marking it as a "requirement" with > mf:requires: > > - mf:requires mf:BasicFederation ; > + mf:feature sd:BasicFederatedQuery ; > > This change is based on mf:requires being used historically as a way > to indicate a test that requires some optional feature of the > implementation (thus making the test optional w.r.t. conformance). > However, we've now split the test suite into conformance requirements > on a per-spec basis (in manifest-all.ttl), and I believe the service > tests should be considered required tests for conformance to the > Federation Extension spec. > > Does anyone have any thoughts on these changes? Both? Running everything and running per-spec make sense. manifest-all.ttl If you want to stress the per-spec natural, let's have one manifest per-spec and make manifest-all.ttl include each area manifest. Andy > > thanks, .greg > >
Received on Thursday, 29 September 2011 11:03:29 UTC