I'd argue that the SCA spec already suggests we should support multiple outputs.

>From Sec. 7.4:

786 – a set of input parameter(s) which match the JAX-WS mapping of
the input parameter(s) of the
787 WSDL operation plus an additional last parameter which is a
ResponseDispatch object typed by
788 the JAX-WS Response Bean mapping of the output parameter(s) of the
WSDL operation, where
789 ResponseDispatch is the type defined in the SCA Java Common
Annotations and APIs
790 specification. [JCA60003]

I think the tricky part is when you try to map (programmatically or
even mentally) from the synchronous Java to the equivalent async Java.
   In that case you have to take apart Holder(s), etc..

But if you start from a WSDL, with one or more outputs... it's easy
enough to map that to both the sync Java and async Java.

Say then we were dealing with a bottom-up, interface.java case, and
wanted to map from the sync Java interface to the async.   We'd need
to first map the sync Java to WSDL, and only then to the async Java.
The relation between sync & async Java might not be as straightforward
as it is for the single output (i.e. where the method return type
simply becomes the ResponseDispatch paramterized type), but I think it
should be natural to extend support for this.

Scott

Reply via email to