Re: RE: RPC Style Issues (3)

> Don't forget that the RPC style is not needed for interop since it does
> not affect the messages that are exchanged. It is a means to convey
> intended programming model and should be designed to be sufficient for
> that purpose.

I disagree with this statement. 

RPC style has two distinct requirements: 

-- Schema requirements that describe what it means to be a conformant WSDL with respect to this style The schema requirements define the structure of the messages, (i.e) whether it should only contain local elements, etc. 

-- Programming hints. 

This is why I made the proposal [1] to divide the rules into two sections. 

You are putting these two requirements as if both talk about programming hints, they don't. The first set of rules define what the schema conformance rules are. 

I would agree with Sanjiva (this is verbatim from my IRC log Sanjiva, i hope you don't disagree quoting you here again ;-)). "Interoperability is quaranteed with the schemas, not the programming model." 

This is why I propose the rules are defined as such, in two separate sections. 

For example, if I were to publish a WSDL with the RPC style URI, but do not follow the rules in the first section, a client who makes use of this style may not be able to process the schema because it is expecting the messages to be formatted in a certain way and they are not. That is why it affects interoperability. I would not know what to do if the complexType contained attributes, or a choice, etc. 

IMHO, David explained it quite well during the telcon. We should be talking about what conformance means. From our perspective, a WSDL is not conformant if it uses the style URI but did not follow the schema rules. Arthur's checker should be able to barf on such a WSDL.  This has nothing to do about how one further processes the message, but is necessary for those who would make use of this style. 

Further, if you want to follow the hint or use DOM to represent your message, conformance can not talk about it unless we get into the business of describing how each respective language such as Java, C++, C# must be handling RPC style. There are specs for that and conformance suites that are tailored. I am sure some members of this community will not be very receptive if we were to get into that business. 

What we are proposing here is the message conformance. Lets not mix it with the intended programming model that is further described in the second set of rules.  

This is the same analogy to using a MEP which is defined to use a message-triggers-fault rule but mistakenly uses faults incorrectly wrt directionality. I say this is a non conformant WSDL, regardless of you know how to implement this MEP or not. 

Cheers, 

--umit


[1] http://lists.w3.org/Archives/Public/www-ws-desc/2003Oct/0057.html