- From: Jeremy Carroll <jjc@hpl.hp.com>
- Date: Wed, 07 Mar 2007 17:42:21 +0000
- To: GRDDL Working Group <public-grddl-wg@w3.org>
We have been discussing introductory text for the tests along the lines of: [[ A GRDDL-aware agent, with security policy set to none, and configured to apply all transforms, produces GRDDL results as shown in these tests. ]] Then, the grddl-on-rdf tests, I think, have only one right answer: the merge (i.e. output3). If we want to stick with three different results for the grddl-on-rdf tests, then the wording above could be, "and configured to apply at least one transform". The content negotiation tests (whether language or mimetype) have three right answers. With the OWL test cases I learnt, after a few comments like Dan's comment: http://lists.w3.org/Archives/Public/public-grddl-wg/2007Mar/0037.html that the description of each test should be as anodyne and bland as possible. e.g. "A test concerning an information resource with two representations, with different mimetypes." And then the text describing multiple results can be something like: [[ A GRDDL-aware agent should output one of the GRDDL results specified for each test. When multiple possible GRDDL results are specified, this is because different GRDDL-aware agents, or GRDDL-aware agents with different policies, may: - retrieve (a) different representation(s) of the information resource - apply, or not apply, different options when converting an XML representation of an information resource into an XPath node set. ]] Of course, this text needs to be consistent with the spec; but by insisting on blandness everywhere else, it should be possible to reduce the risk of inconsistency on a test-by-test basis. Jeremy
Received on Wednesday, 7 March 2007 17:42:56 UTC