xsi:schemaLocation and XSD 1.1 schema composition

[brunato]

Implementing a dynamic schema load for xmlschema package i found a set of test cases related with bug 4078.

This bug differentiates XSD 1.0 and XSD 1.1 about the logic of schema composition, so test cases addB156 and addB158 were marked as valid for XSD 1.1. But test cases addB159, addB162, and addB166 are still marked as invalid also for XSD 1.1.

I would ask if this is an oversight or if an additional discussion is needed before changing the state of these cases.

About XSD 1.1 schema composition as described in the 4th item of G.1.15, I'm unsure about the new rule described by this sentence:

Note, however, that if processing such an xsi:schemaLocation attribute causes new components to 
be added to the schema, then the new components cannot change the assessment outcome of 
any information items already seen before the element bearing the xsi:schemaLocation attribute.

What does it means? That I have to validate again the XML document considering the new added components? Or what else?

Thank you

[michaelhkay]

I have the same problems with these test cases. See my analysis at https://saxonica.plan.io/issues/6954

Bug 4078 is now at https://www.w3.org/Bugs/Public/show_bug.cgi?id=4078. This bug led to tests addB156 and addB158 changing. The failure to do the same for tests such as addB161 was because I failed to notice at the time that they had the same problem, and therefore didn't include them in that bug report.

Like you, I have no idea what the note that you cite is intended to mean, but for Saxon when processing XSD 1.1 I am now ignoring xsi:schemaLocation and xsi:noNamespaceLocation attributes for namespaces that have previously been encountered.

[michaelhkay]

Actually, I queried a slighly different set of test cases.

You queried addB159, addB162, and addB166 which today are marked as queried under bug 4078. The status of queried means there's an open bug against the tests, so the expected results shouldn't be considered reliable. I'm being cautious here, in the absence of a higher authority I feel qualified to mark a test as queried, but not to change the expected results unilaterally. My own test driver essentially ignores a test if it is marked as queried.

In my Saxonica bug I queried addB161 and addB164. I shall change the metadata for those tests accordingly.