rick rineholt wrote:
> There is also the practice that meta data is being
> added to WSDL (i.e.
> xmlns:sdojava="commonj.sdo/java ..
> sdojava:package="foobar.fake.com") This *may*
> be unique to client or server. I don't like this
> practice but I suspect will
> see more in the future.
>
This raises an interesting point about people extending documents with
proprietary extension elements as general practice. I can see scenarios
where it is essential that we read a specific annotated instance rather
than use one from a central registry.
There has been some unconcluded discussion in the spec process about
supporting the use of wsdlLocation (from 2.0) to define the actual WSDL
instance to use. I think Rick's note shows that we need to support that
on every element that references a WSDL so that modifications applying
to that use can be picked up.
This would allow us to support WSDL import without any modification to
the spec schema. Something like:
<module name="foo"
<!-- per module definition -->
wsdlLocation="http://foo.org foo.wsdl">
<entryPoint name="ep1">
<interface.wsdl
interface="http://bar.org#wsdl.interface(Bar)"
<!-- per usage definition -->
wsdlLocation="http://bar.org bar.wsdl"
/>
<binding.ws
portType="http://bar.org#wsdl.endpoint(Boo)"
wsdlLocation="http://bar.org boo.wsdl"
/>
...
<componentType name="baz">
<service name="s1">
<interface.wsdl
interface="http://baz.org#wsdl.interface(Baz)"
wsdlLocation="http://baz.org baz.wsdl"
/>
...
Things would default outward if not specified - per-use to per-module to
versions in the default runtime.
--
Jeremy
