It is VERY unfortunate that the interface needs information from both the portType AND the binding. This is a topic of discussion for WSDL version 2.
JAX-RPC says (section 4.3.3): "The name of the Java interface is mapped from the name attribute of the wsdl:portType element. ... If the mapping to a service definition interface uses elements of the wsdl:binding ..., then the name of the service definition interface is mapped from the name of the wsdel:binding element." Note the name of the spec. It contains the string "RPC". So this spec, and WSDL2Java, assumes that the interface generated from the portType is an RPC interface. If information from the binding tells us otherwise (in other words, we use elements of the wsdl:binding), then the name of the interface is derived instead from the binding. Note that we could have one portType - pt - and two bindings: bRPC and bDoc. Since doc/lit changes what the interface looks like, we cannot use a single interface for both of these bindings, so we end up with 2 interfaces - pt and another named bDoc - and two stubs - bRPCStub (which implements pt) and bDocStub (which implements bDoc). Ugly, isn't it? But you can see why it's necessary. Since doc/lit changes what the interface looks like, and we could have more than one binding refering to a single portType, we have to create more than one interface and each interface must have a unique name. Russell Butek [EMAIL PROTECTED] Thomas Sandholm <[EMAIL PROTECTED]> on 03/13/2002 05:24:30 PM Please respond to [EMAIL PROTECTED] To: [EMAIL PROTECTED] cc: Subject: WSDL2Java doc/literal port type interface I have noticed that the newest code from cvs generates a port type interface using the wsdl:portType name when rpc/encoded is used, but if doc/literal is used the wsdl:binding name is used. Is there any particular design rationale behind this? I think it seems to break the idea of not exposing the fact that doc/literal is used to the client or to the service implementer. Besides that it seems like this interface should not have any knowledge about the binding, so it seems strange that it uses the name of the binding in its own name in the doc/literal approach. Any thoughts about this? /Thomas
