Why?

The rules for converting a time to a string are defined in F+O section 18.2, and it is quite clear that the UTC timezone should be represented as Z.

True. I ignored the fact that the input of these queries is also dependent on a schema, so I would suggest adding yet another schema dependency to the test case definition.

I'd suggest strongly that rather than wait for busy volunteers to change all the tests to add schema dependencies, you do what the rest of us are doing, which is to infer a schema dependency from the presence of a source document that needs validating.

I am aware that it takes valuable time to commit the test case fixes, and I very much appreciate the work (mostly) done so far by you, O'Neil, and Tim. If everyone agrees, I’ll be glad to check in obvious fixes by my own.

Regarding the validation of input documents, I observed that many test cases will yield correct results even if an attached schema is not evaluated. This is why I looked for the few remaining tests that have no explicit schema dependency. Otherwise, I believe, a test suite driver for an implementation without schema support would either have to ignore all tests that have any schema declaration included – which would be a pity – or try to evaluate a query, and check for a schema declaration if the query fails, and then assume that it failed because of schema support – which doesn’t seem too elegant. If there is another solution that I have omitted, and that is a better solution that explicit schema dependencies, feel free to tell me.

I am sorry that the additional declarations have caused quite some extra effort, but I assume that there is now hardly any test left that lack an explicit schema dependency.

 Closed (and added a comment to #21238).