Since this thread has died down, I decided to checkin the incremental
change I made inspired by this discussion, into 2.x in r1070926 (and
1070928).  I mentioned this in my comment on TUSCANY-3832, which was
an issue I found in trying to make this fix.

In terms of comprehensibility, my patch improves things by creating a
new "notSubjectToWrapping" boolean on Operation, for use in the
Input2Input/Output2Output transformers in place of where we'd been
making the "sourceBare/targetBare" calculation.   I chose
"notSubjectToWrapping" rather than "subjectToWrapping" so we could
have the more convenient default of "false".

This requires an operation to mark itself as "notSubjectToWrapping"
rather than relying on the absence of a wrapper (I fixed the JAXWS
intf processor accordingly), and in doing so it enables Simon's
bare->wrapped WSDL transform to work.

If we want to revisit and try to further simplify things.. I'll try to
at least keep up with the discussion !
--------

Like I mentioned, this particular test (since moved to
'contrib/samples' which I updated in r1070929), exposes another issue
(TUSCANY-3832), which is how to bypass the Mediator.copy in the
binding-sca-runtime implementation in cases when the
DataTransformationInterceptor will be established on the invocation
chain.    I didn't come up with too great a solution.... but I at
least avoided my earlier idea of using a message header... and did
something which is I think more in line with things we've done to
date.

By the way, the interface configuration on the component reference
actually has no effect (Simon you may have thought it did).   The
"target" operation (i.e. the reference binding contract) actually
comes from the target service "componentType" interface contract in
this case... the interface you specified had no effect, so I removed
it from the test.

Scott

Reply via email to