On Wed, 2005-08-24 at 09:19 -0400, Glen Daniels wrote: > Dims, you can only do that if WSDL says that's what you do (otherwise > it's not interoperable, or correct according to schema). > > I don't see the problem Sanjiva is worried about, though. If the > elements are unqualified, then the references are simply unqualified as > well... <input element="foo"/> instead of <input element="ns:foo"/> - > what's the big deal?
The rules for resolving qualified *values* that WSDL 1.1 assumed came from XSLT, not from WSDL. (See http://www.w3.org/TR/xslt#qname.) So while you're correct about how one could interpret element="foo" (because XSDL explicitly states that default NS doesn't apply), it was not the intent of WSDL 1.1 that unqualified names would be used as values of @element and @type. Furthermore, WSDL 1.1, does not specify or imply the behavior you propose. For those of you who care about the WS-I Basic Profile, it clarifies this issue too: http://www.ws-i.org/Profiles/BasicProfile-1.0-2004-04-16.html#refinement16552904 However, the solution they advocate (which is what the WS-RF schema uses) is actually incorrect w.r.t. XSD IMO .. the EIIs and AIIs defined in the schema which is imported inside another schema are not visible outside of the enclosing schema. So in other words, WSDL would not be able to see them. WS-I BP gets around that by defining its own rules .. but technically that's not correct. Anyway, I take back my previous comment: this *specific style of XSDs* (per the WS-I BP clarification) should be supported as a special case. The style being of course: <xsd:schema> <xsd:import namespace=".." location=".."/> </xsd:schema> This is analogous to what we allow in WSDL 2.0 by allowing xsd:import directly as a child of wsdl:types. We clarified this whole issue in WSDL 2.0 after lots of thought. See: http://www.w3.org/TR/2005/WD-wsdl20-20050803/#xsd-types There we say (in 3.1.1.1): "The namespace attribute information item defines the namespace of the element declarations and type definitions imported from the referenced schema. The referenced schema MUST contain a targetNamespace attribute information item on its xs:schema element information item and the values of these two attribute information items MUST be identical. It is an error to import a schema that does not have a targetNamespace attribute information item on its xs:schema element information item. Such schemas must first be included (using xs:include ) in a schema that contains a targetNamespace attribute information item on its xs:schema element information item, which can then be either imported or inlined in the WSDL 2.0 document." === In WSDL 1.1, the trick we used was to say you can use WSDL 1.1's <import> element to import XSDs. (See Section 2.1.2 of http://www.w3.org/TR/wsdl.) So the right way to do the above would've been to write: <wsdl:definitions ...> <wsdl:import namespace=".." location=".."/> .. </wsdl:definitions> The WSDL processor was responsible for sniffing the stuff at the location URI and figuring out whether to import it as XSD or as WSDL (or some other type system .. in the WSIF context people have used it to import weirdities like COBOL copy books ;-)). However, WS-I BP folks, in their infinite wisdom, felt this was no good and declared this illegal: http://www.ws-i.org/Profiles/BasicProfile-1.0-2004-04-16.html#refinement16498504 (It baffles me how WS-I BP folks felt they could re-write the WSDL 1.1's recommended authoring style (http://www.w3.org/TR/wsdl#_style) but that's what they did. Oh well.) Sanjiva.
