Re: WSDL 1.1 schema question

Arthur Ryman wrote:

>
> Sanjiva,
>
> I agree that schema ain't good enough. The spec expresses many 
> additional semantic rules that are beyond the scope of XSD. For 
> example, AFAIK, XSD can't express semantics involving imported WSDL 
> documents.
>
> I would like to see the WG define a schema that expresses as much as 
> possible and give that schema "official" status. That is what I meant 
> by the schema being normative, i.e. that it is an official part of the 
> spec. The schema is not exhaustive though, since it can't express 
> everything.
>
> The approach for the extra-schema semantics should be that the WG 
> include test assertions in the spec. A WSDL validator could then 
> implement the test assertions. The combination of the normative XSD 
> plus all the test assertions should define what it means for WSDL to 
> be WSDL-valid.
>
> We therefore have a sequence of checks we can perform on WSDL:
>
> 1. Syntax - well formed XML
> 2. Structure - XSD valid wrt to the normative wsdl.xsd
> 3. Semantic - all test assertions in the spec are satisfied
>
> So to really acheive the needed level of precision, we need both a 
> normative XSD and a normative set of test assertions.

I agree with Arthur. From a tools perspective, if the schema was 
normative, a WSDL "processor" could rely on the a schema validating 
parser to do most of the work. The rest can be inferred based on a 
structurally sound WSD document by further analysis.

If  people are worried about most of the structural normative aspects 
appearing only in the schema, I think we should still publish the 
assertions related to the structure, but clearly indicating that they 
appear in the normative schema and the schema is the final word. It is 
easier for us (or the commentors) to debug the the wording of the 
assertions when we know what is covered in the schema and what is not.

I do realize that we will not be able to represent everything in the 
schema, but this will be helpful to tools. The assertions will also 
clearly identify what needs to be verified after schema validation.

--umit

>
>
>
> Arthur Ryman
>
>
>
> "Sanjiva Weerawarana" <sanjiva@watson.ibm.com>
> Sent by: www-ws-desc-request@w3.org
>
> 03/26/2003 10:02 AM
>
>        
>         To:        "Jeffrey Schlimmer" 
> <jeffsch@windows.microsoft.com>, Arthur Ryman/Toronto/IBM@IBMCA
>         cc:        <www-ws-desc@w3.org>
>         Subject:        Re: WSDL 1.1 schema question
>
>        
>
>
>
>
> I agree in concept, of course. However, in practice the XSD would
> have to have way more flexibility than really legal due to XSD
> limitations. Those would have to be documented in annotations,
> meaning that XSD validation simply ain't good enough.
>
> In that setting, I'm not so sure making the XSD normative is
> very useful or accurate.
>
> What does our resident schema expert think?
>
> Sanjiva.
>
> ----- Original Message -----
> From: "Jeffrey Schlimmer" <jeffsch@windows.microsoft.com>
> To: "Arthur Ryman" <ryman@ca.ibm.com>
> Cc: <www-ws-desc@w3.org>
> Sent: Wednesday, March 26, 2003 1:03 AM
> Subject: RE: WSDL 1.1 schema question
>
>
> >
> > +1
> >
> > > From: Arthur Ryman [mailto:ryman@ca.ibm.com]
> > > As a tool developer, I find the second class status of the schema to
> > be a
> > > bad idea. After all, schema is a precise formal language that supports
> > > machine processing. My goal as a WG member is that we should provide a
> > > normative schema since that is unambiguous in comparison with the
> > narative
> > > prose in the spec.
>
>
>

-- 
Umit Yalcinalp					
Consulting Member of Technical Staff            400 Oracle Parkway
Oracle
Phone: +1 650 607 6154                          Redwood Shores,
Email: umit.yalcinalp@oracle.com                CA 94065, USA