RE: Issue 51: Should soap:header@part be nmtokens?

(this message proposes a reformulation of issue 51)

 Kevin,
 parts of messages are not reused because they just simply 
specify a type or element - the type and element definitions can 
be reused. So your 2 cannot be done and it's not a bug, it's a 
feature.

 As for header/@parts, the difference between the Body and a 
Header is that the Body may contain multiple XML elements, while 
a header IS AN XML element. What you are asking for may be a 
facility to specify groups of headers, but I personally don't 
think this is necessary - it doesn't make sense to group headers 
which will not be grouped in the actual messages anyway.

 As for multiple headers defined by parts of the same message, I 
don't see the need for consolidating them into one line - there 
would be the artificial difference from the case of multiple 
headers defined by parts of different messages. So what you want 
would just be a syntactic sugar - addition to the syntax, no 
change to the functionality; and I don't think it is necessary in 
this case.

 I do agree the message attribute on soap:header may be made 
optional; and I'd even prefer removing the attribute altogether - 
making header parts belong to the same message as body parts.
 In fact, this may be the assymetry between soap:body and
soap:header that issue 51 should be concerned about (here is the
proposed reformulation of issue 51): "soap:body doesn't specify a
message while soap:header does".

 My position is to remove the attribute 'message' from 
soap:header as it breaks the layering.

 Best regards,

                   Jacek Kopecky

                   Senior Architect, Systinet Corporation
                   http://www.systinet.com/



On Tue, 9 Jul 2002, Liu, Kevin wrote:

 > 
 > Hi Jean-Jacques,
 > 
 > I totally agree that the purpose and process order of the header blocks and the body blocks are different by design. But the question is more about soap:body@parts v.s soap:header@part. Maybe an illustrative example will make it clear.
 > 
 > Say we define two messages for user info,
 > 
 >     <message name="msgLogin">
 >         <part name="userID" type="xsd:string"/>
 >         <part name="password" type="xsd:string"/>
 >         <part name="language" type="xsd:string"/>
 >     </message>
 > 
 >    <message name="msgProfile">
 >         <part name="userID" type="xsd:string"/>
 >         <part name="Country" type="xsd:string"/>
 >         <part name="Language" type="xsd:string"/>
 >         <part name="preferredCustomer" type="xsd:boolean"/>
 >         <part name="photo" type="some binary type"/>	
 >     </message>
 > 
 > 
 > Just fake a portType,
 > 
 >      <portType name="UserManagement">
 >         <operation name="login" >
 >            <input message="tns:msgLogin"/>
 >            <output message="xsd:string"/>
 >         </operation>
 > 
 > 	...
 >     </portType>
 > 
 > Now say we want to bind this portType to SOAP, and just for illustration purpose, we want userID and password in soap:body, userID and Language in soap:header. Given the current WSDL construct, I may need to do the following: 
 >  
 >     <binding ...>
 >         <soap:binding style= ... transport="http://schemas.xmlsoap.org/soap/http"/>
 >         <operation name="login">
 >            <soap:operation ...>
 >            <input>
 >                <soap:body parts = "userID password" ..../>
 >                
 > 	   <soap:header message = "msgLogin" part = "userID" ..../>
 >                <soap:header message = "msgLogin" part = "Language" ..../>	
 >            
 > 	</input>
 > 	...
 >         </operation>>
 >     </binding>
 > 
 > My questions:
 > 
 > 1. For the soap:header section, doesn't it make more sense to
 > just use one line when all the parts are from the same
 > message, symmetric to the case for soap:body?
 >            <soap:header message = "msgLogin" PARTS = "userID Language" ..../>
 > 
 > 2.  In the above example, I may define a "registeration"
 > operation, and I want its input contain some parts from
 > msgLogin (say userID and password) and some parts from
 > msgProfile (say country and Photo). How can I do that with the
 > current portType construct? Note I don't want to combine this
 > two messages for this purpose since each of them may be reused
 > by a couple of other operations.
 > 
 > 3. If 2 can't be done, does it make sense to allow soap:body
 > contains info define in more than one messages? currently it's
 > not possible to say,
 >                <soap:body message = "msLogin" parts = "userID password" ..../>
 >                <soap:body message = "msProfile" parts = "country photo" ..../>
 > 
 > 	If this is a valid case, the @message for both body and
 > header should be optional and default to the enclosing
 > input/output message (WSDL1.1 grammar indicates that
 > soap:header@message is required, bullet 1 above hopefully
 > makes it obvious that optional makes more sense)
 > 
 > Regards,  Kevin
 > Technology Architecture Group, 
 > SAP Labs, Palo Alto
 > 650-849-5167
 > 
 > 
 > -----Original Message-----
 > From: Jean-Jacques Moreau [mailto:moreau@crf.canon.fr]
 > Sent: Tuesday, July 09, 2002 2:28 AM
 > To: Liu, Kevin
 > Cc: 'Jonathan Marsh'; www-ws-desc@w3.org
 > Subject: Re: Issue 51: Should soap:header@part be nmtokens? (was RE:
 > ACTION: All wg members to review editorial issues by July 11)
 > 
 > 
 > I think the onus is on SOAP. SOAP 1.2 clearly differentiates between SOAP header blocks and the SOAP body.
 > 
 > A SOAP header block is typically additional data that helps the ultimate SOAP receiver process the SOAP body, or data targeted to a SOAP intermediary along the SOAP message path. Syntactically, a SOAP header is contained within the SOAP Header element.
 > 
 > The SOAP body carries the main intent of the SOAP message. It is (implicitely) targeted to the ultimate SOAP receiver. It has an application defined structure, not one defined by SOAP.
 > 
 > Also, there are differences in the way SOAP header blocks and the SOAP body are processed. In particular, the SOAP body is not processed unless all mandatory SOAP header blocks are recognized and understood by the SOAP processor.
 > 
 > So, overall, I think we just have to recognize they are enough syntactic and semantical differences, that we have to keep them separate in WSDL 1.2 as well. (And, BTW, I think it would be wrong to dissociate from the direction set by the XMLP WG, unless one really think they got it wrong, in which
 > case we should raise it as a
 > Last Call issue for SOAP 1.2.)
 > 
 > I hope this helps,
 > 
 > Jean-Jacques.
 > 
 > [1] http://www.w3.org/TR/2002/WD-soap12-part1-20020626/#procsoapmsgs
 > 
 > "Liu, Kevin" wrote:
 > 
 > > WSDL1.1 has different constructs for body and header, but doesn't provide any justification about the differences: [...]
 > > Given the fact that soap:body is just a particular kind of soap:header, it's not clear why wsdl should treat them so differently, in particular: [...]
 >

Received on Monday, 15 July 2002 09:07:35 UTC