Re: issue: embedding semantic descriptions?

Jacek,

All three options you outlined are viable options and theoretically I don't
see a problem with supporting them all. But this raises a concern regarding
how flexible the spec should be. If it is so flexible and lets you do the
same thing in 10 different ways, it is very confusing. This may also open
up cases where annotations get out of synch. What if someone creates
external annotations and embedded annotations in the same language for the
same concept but they are not consistent? Then, we have to specify all
sorts of conflict resolution rules that cannot be enforced. I think from a
practical usability point of view, it is better to restrict choice to keep
things simple.

Since we cannot prevent people from using the option 1 that you have
outlined, we probably should note it as a recognized practice in the
document but not go beyond that to keep things simple.

Regards
Rama Akkiraju

public-ws-semann-request@w3.org wrote on 04/17/2006 12:03:36 PM:

>
> Hi, to bring one possibly contentious issue in the open, I suggest that
> we consider whether we want to allow embedded semantic descriptions or
> whether we want to restrict SAWSDL to just URI references to external
> things?
>
> Currently we have modelReference and schemaMapping that contain a URI.
> I see a number of additional options that are not necessarily exclusive:
>
> 1) we can document that these URIs may point to elements within the WSDL
> file with a particular ID, so we could for example put a model or an
> XSLT stylesheet inside wsdl:description, put an xml:id on it and then
> refer to that ID with a fragment identifier in a URI.
>
> <wsdl:description ...>
>   <xsl:stylesheet xml:id="transformation">
>     ...
>   </xsl:stylesheet>
>   <wsdl:types>
>     <xs:schema ...>
>       <xs:element sawsdl:schemaMapping="#transformation" .../>
>     </xs:schema>
>   </wsdl:types>
>   ...
> </wsdl:description>

>
> 2) we can provide a container element akin to wsdl:types that would
> contain semantic descriptions defining some URIs that could then be the
> values of modelReference
>
> <wsdl:description ...>
>   <sawsdl:models>
>     <wsml:wsml ...>
>       <wsml:concept id="http://example.com/foobar"/>
>     </wsml:wsml>
>   </sawsdl:models>
>   <wsdl:types>
>     <xs:schema ...>
>       <xs:element sawsdl:modelReference="http://example.com/foobar" .../>
>     </xs:schema>
>   </wsdl:types>
>   ...
> </wsdl:description>

>
> 3) we can provide an element equivalent to the modelReference attribute
> (and similarly for the schemaMapping attribute) that would contain the
> actual semantic description (or transformation) without the necessity of
> giving it any URI.
>
> <wsdl:description ...>
>   <wsdl:types>
>     <xs:schema ...>
>       <xs:element ...>
>         <sawsdl:schemaMapping>
>           <xsl:stylesheet>
>             ...
>           </xsl:stylesheet>
>         </sawsdl:schemaMapping>
>       </xs:element>
>     </xs:schema>
>   </wsdl:types>
>   ...
> </wsdl:description>


> Note that option 1 requires no new elements or attributes from us, it
> would be just documentation for recognized practice. However, option 2
> could be more compatible with the WSDL 2 component view of the
> documents. Finally, option 3 is really here just for consideration in
> case somebody else is interested in pursuing it.
>
> What do you think?
>
> Jacek Kopecký
>
> --
> Digital Enterprise Research Institute
> University of Innsbruck, Austria
> Phone: +43 512 5076481
> Org:   http://www.deri.org/
> Blog:  http://jacek.cz/blog/
>
>
>

Received on Tuesday, 25 April 2006 04:01:35 UTC