NEW ISSUE: Semantics of wsa:UsingAddressing@wsdl:Required="false" case are unclear.

Title: Semantics of wsa:UsingAddressing@wsdl:Required="false" case are 
unclear.

Description:

The intent of this case is that the endpoint should support both 
existing behavior and that defined in the WSA core.  However, it is not 
clear how the endpoint will tell which case is in effect.  The WSA MAPs 
are abstract SOAP properties, and as such it is up to the binding to 
decide how to put them on the wire.  For example, the (normal) presence 
of an http action header implies the presence of the wsa:Action 
property.  Similarly, in an email binding, the from:, and to: headers 
may carry the [source endpoint] and [destination] MAPs.  There appears 
to be no general way to inspect an incoming SOAP message and know for 
sure whether the sender intended to include MAPs.  Given this, it is 
unclear whether to flag an incomplete set of MAPs as a bad message sent 
by a WSA-aware sender or as a normal message sent by a non-WSA-aware sender.

Justification:

The WSDL binding should be able to handle both existing behavior and 
WSA-aware behavior without requiring multiple endpoint definitions.

Target: WSDL binding

Proposal:

The rules described in [1] are compatible with both existing clients and 
WSA-aware clients following the rules described in the WSA core.  They 
do not require a determination of whether the client meant to use WSA.  
If they are applied in the case that 
wsa:UsingAddressing@wsdl:Required="false", then

    * Existing WSDLs work as-is
    * Endpoints with wsdl:Required="true" follow the strict semantics
      defined in the core.
    * Endpoints with wsdl:Required="false" follow the proposed rules and
      are clearly compatible with either kind of client.

Note that there are several possible variants on these rules which 
maintain compatibility.  In particular:

    * The proposed "return to sender" default can be omitted.
    * Anonymous endpoints may be allowed.