Re: [wsbpel] Issue - 152 - Proposal to Vote

["Yaron Y. Goland" <ygoland@bea.com>]

wsdl:endpoint was never designed to be an EPR. That WS-MD has chosen to 
allow wsdl:endpoint to be used as an EPR is unfortunate but I can see no 
compelling reason why we should complicate the BPEL spec to make up for 
WS-MD's unfortunate design decision.

Therefore I continue to propose that we remove the reference-scheme 
attribute and respectfully suggest to the WS-MD crowd that they wrap 
wsdl:endpoint in a container element.

	Thanks,

		Yaron

Ron Ten-Hove wrote:

> 
> 
> Yaron,
> 
> Yaron Y. Goland wrote:
> 
>  > Since service-ref is an attribute how can reference-scheme be an
>  > attribute of service-ref? Attributes in XML cannot have attributes.
> 
> This is a typo / misunderstanding; service-ref is an element, with a
> reference-scheme attribute. Service-ref serves as a "wrapper" for the
> EPR, so it must be an element.
> 
>  >
>  > Furthermore there can never be an ambiguity about the namespace of the
>  > EPR because the EPR root element MUST always declare what its
>  > namespace is, even if it's nothing.
>  >
>  > So I'm still unclear as to what role the service-ref plays.
> 
> The namespace of the enclosed EPR may be insufficient to identify the
> addressing scheme in use. For instance, WS-MD uses wsdl:endpoint
> elements for EPRs, and it is conceivable that other addressing schemes
> may follow the same approach (if only newer versions of WS-MD!).
> 
> -Ron
>