Re: [wsrf] WS-RMD issue: precedence of RMD location

[Daniel Jemiolo <danjemiolo@us.ibm.com>]

Hi Tim,

Here are some possible answers:

> Should we add a statement to say that the MDD
referenced from 
> the WSDL and the one referenced from the resourceProperty
SHOULD 
> be the same Qname?

I think that the two MetadataDescriptor names MUST
be the same; once 
the RMD document is resolved, a resource should point
to one descriptor 
in the document no matter how the client came to find
it.

The URI of of the RMD document is another story -
the two URIs should 
resolve to the same document, but one (in the WSDL
portType) may be 
relative.

> Does the resourceProperty takes precedence if
they aren't the same?

I think that remote clients should only use the property
when trying 
to locate the document and descriptor. This seems
like the only way 
to guarantee robust metadata exchange (without reinventing
WS-MEx). 

> This would provide some wiggle room for services that start out 
> with no MDD, buthen get one and might like to
keep happy those 
> clients who used the original WSDL.

Since RMD support is not part of a resource
type definition (like 
WSDM capabilities, for example), I think
that this case is allowed 
by the current proposal. If/when a MDD
is applied to a resource, the 
location/name of the MDD MUST be added
to the WS-RP document and 
the WSDL portType. 

Clients using the original WSDL should
not encounter problems 
because of the availability of the new
property (right? I might have 
misunderstood the scenario).

Dan

Tim Banks <tim_banks@uk.ibm.com>

03/15/2006 09:32 AM

To	Daniel Jemiolo/Durham/IBM@IBMUS
cc	wsrf@lists.oasis-open.org
Subject	Re: [wsrf] WS-RMD issue: precedence of RMD location

Hi Dan,

Yes - this would be an improvement....

Daniel Jemiolo <danjemiolo@us.ibm.com> wrote on 08/03/2006 21:51:56:

>
> During the last TC call, Dave S. brought up the fact that WS-RMD has
> two ways of specifying the location of the resource's RMD but no
> order of precedence for reading/verifying them. What follows is a
> proposed resolution:
>
> 0. If a resource has an RMD, it must have the location specified in
> a resource property and the WSDL portType attribute. Today, the
> property is optional.

Your suggestion makes sure that an MDD is present in the Resource
Properties.
Should we add a statement to say that the MDD referenced from the WSDL
and
the one referenced from the resourceProperty SHOULD be the same Qname?
Does the resourceProperty takes precedence if they aren't the same?
This would provide some wiggle room for services that start out with no
MDD, but
then get one and might like to keep happy those clients who used the
original WSDL.

>
> 1. The property value must be an absolute URI. It should be easy to
> generate this value at runtime, when the final deployment info of
> the resource is known. The value provided in the WSDL attribute may
> be a relative path, and this may not resolve for remote clients (but
> may be useful during design time).
>
> Note: Relative paths may not resolve because many SOAP engines use
> virtual paths to distinguish between different web services. A
> client that discovers the WSDL through ?wsdl query or WS-
> MetadataExchange would not know how to map the relative path against
> the URL of the service.

+1

>
> 2. Simplify the data type of the WSDL
> portType/@metadataDescriptorLocation attribute from "list of
pair of
> URIs" to "xsd:anyURI" so that there is one descriptor
per portType.

+1

>
>
> Thoughts?
>
>
>
> Dan Jemiolo
> IBM Corporation
> Research Triangle Park, NC
>
>
> +++ I'm an engineer. I make slides that people can't read. Sometimes
> I eat donuts. +++