What is the SOAP MEP for In-only

Folks, 

It seems that we have a hole in our specification. 

We were trying to specify a WSDL interface that utilizes a single input
message, i.e. with In-only MEP. 

Currently, per section 2.9.2 of Part3 [1] states the following:  

{
{soap mep}, a wsdls:anyURI, which is an absolute URI as defined by [IETF
RFC 2396], to the Binding Operation component. The value of this
property identifies the SOAP Message Exchange Pattern (MEP) for this
specific operation. If no specific value is assigned, then the value
assigned by the default rules apply (for SOAP 1.2, see 2.10.3 Default
Binding Rules). It is an error for this property to not have a value
(which MAY happen if the default rules are not applicable).
} 

We then looked at (Section 2.10.3 Default Binding Rules) which says: 

{
SOAP MEP Selection. If the Interface Operation component's {message
exchange pattern} property has the value
"http://www.w3.org/@@@@/@@/wsdl/in-out" then the default value of the
{soap mep} property for the corresponding Binding Operation component is
"http://www.w3.org/2003/05/soap/mep/request-response/" identifying the
SOAP Request-Response Message Exchange Pattern as defined in [SOAP 1.2
Part 2: Adjuncts]. If the Inteface Operation component has any other
value for the {message exchange pattern} property, then no default value
is defined for the {soap mep} property of the corresponding Binding
Operation component.
}

This means that for an In-only MEP, we have no default SOAP MEP.
However, it is an error for the property not to have a value. 

This means the user has two choices: 

-- Define a new SOAP MEP for an In-only MEP. Since there is no
recommended SOAP input MEP that is defined in the SOAP Adjuncts, that is
the only choice for the user. Right? 

-- Use a predefined SOAP MEP, such as request-response. But, this does
not really match the abstract definition of the operation, as there will
be a response. This means then I will have to change my interface to use
a different MEP, that allows a response. Robust-In? 

What are we really suggesting the community to do in this case? 

This is rather disturbing to us. It is well known that some of our MEPs
may not have corresponding SOAP MEPs or appropriate bindings that may be
addressed by extension mechanisms, but In-only is a common WSD MEP that
the community uses already. 

Did we miss a decision pertaining to this case? We don't think that
expecting the users to define a new SOAP MEP is acceptable for interop
purposes as currently the only MEP that is defined with the bindings is
In-Out...

Thanks. 

--umit and kevin

[1]

Received on Friday, 14 January 2005 22:35:04 UTC