Re: GRDDL Going to Last Call: Relevance to SAWSDL

Jacek Kopecky wrote:
> Dear Harry, 
> thanks for letting us know about the upcoming GRDDL LC. 
>   
>> 2) GRDDL applies to XML documents in general, including XHTML and XML
>> Schema. Insofar as "lifting" as defined in SAWSDL by
>> "liftingSchemaMapping" refers to XML Schema, GRDDL is a superset of
>> "lifting".
>>     
>
> I'm afraid there might be a misunderstanding stemming from our attribute
> name - schema mapping - because we point to transformations that
> transform data, e.g. XML to RDF, not schemata, e.g. XML Schema to OWL.
> Do you think a renaming to something like liftingDataMapping and
> loweringDataMapping would help?
>   
It's been a while since I've looked at SAWSDL, but no - there's no
reason to change "liftingSchemaMapping" to "liftingDataMapping." I don't
think I misunderstand, as I thought the purpose of XML Schema-based
lifting and lowering was the conversion of an XML document itself, *not*
the XML Schema, to RDF via mappings embedded in the XML Schema, as
defined by SAWSDL.  So, one gets an XML instance document, one schema
validates with a SAWSDL-enabled schema, gets a PSVI, and therefore one
also gets a RDF version of the XML-instance document. GRDDL is a
differentl way of doing that by putting some (possibly Turing-complete)
black-box transformation in either the document itself or the XML Schema
or RDF/S document at its namespace.
>> 3) Lastly and most importantly, SAWSDL specifIes a way of specifying the
>> mapping that is componentized to the types of the XML Schema, while
>> GRDDL currently only operates over the whole XML document, and so cannot
>> at this time specify transformations at the level of simple and complex
>> XML Schema types.
>>     
>
> Our mappings also work on the whole document - only the
> liftingSchemaMapping that is on the element declaration or type
> definition for the root element of the message gets used.
>   
That would be very similar to GRDDL, in fact, pretty close to the same
thing as GRDDL applies to whole documents. Where SAWSDL has capabilities
GRDDL doesn't is when it applies to element declarations or type
definitions that *aren't* on the root document. Where GRDDL does
something interesting I think SAWSDL doesn't it's its ability to embed
transforamtions in profiles and namespace documents.
>> I think these observatons require no changes to SAWSDL as the spec
>> stands. However, it would be nice to put in a reference to GRDDL, such
>> as after the sentence: "Other languages, such as XQuery can also be
>> used." "GRDDL (Gleaning Resource Descriptions from Dialects of Language)
>> is a way of "lifting" XML documents, including XML Schema, to RDF.
>> Transformations created for use by GRDDL can be used as lifting mappings
>> for SAWSDL if appropriate."
>>     
>
> If I understand it correctly, GRDDL puts links to the mappings in the
> data itself, so we would not really have much to point to from the data
> schema. We can say that instead of pointers in the WSDL (in the schema),
> one could employ pointers in the data itself, but I suspect this
> wouldn't fly with Web Services people because it would be repeated
> overhead.
>   
Either in the data itself or the namespace document (which could be an
XML Schema document, as pointed out in the spec) or profile. Putting
pointers in the namespace document would not be repeated overhead. 
Still, obviously SAWSDL and GRDDL have different use-cases: GRDDL for
transforming XML dialects and XHTML dialects to RDF on the client side, 
and SAWSDL for the much more Schema-based world of Web Service
integration, and I think both are very useful and pragmatic standards.
> So in the end, a link to GRDDL in the SAWSDL spec might not be all that
> useful, but perhaps GRDDL might be mentioned in our usage guide.
>
>   
>> In other words, nothing new should be implemented, and the note merely
>> notices the relationship and that if someone makes a GRDDL transform for
>> an XML Schema to RDF someone can save effort and use the same transform
>> if wanted in a SAWSDL. I would also like to add a note about the
>> relationship to SAWSDL one of the GRDDL documents.
>>     
>
> Since SAWSDL doesn't define any mappings itself, I suspect that GRDDL
> could only say something like "instead of linking the mappings in the
> data itself, they may be linked from the schema, e.g. using SAWSDL's
> liftingSchemaMapping, however, this particular mechanism does not allow
> combining multiple mappings."
>   
Again, I think the major difference between GRDDL and SAWSDL is that
1) SAWSDL requires XML Schema
2) And therefore SAWSDL can do type and element level transformations.

For those languages that do not have XML schemas and that do not require
type and element-level transformations, it makes sense to use GRDDL.
Otherwise, it makes sense to use SAWSDL, and since WSDL Web Services are
heavily Schema-based, yes - of course SAWSDL makes more sense.

You've made great sense Jacek, and I'm glad to see this progress. I
think SAWSDL actually touches upon an issue that the GRDDL WG postponed
[1], i.e. how to push transformations down to the element level.
Hmmm....makes me think a combined GRDDL-SAWSDL processor might be an
interesting project for someone to take on.
> Does it make sense? 8-)
> Best regards,
> Jacek
>
>   
[1] http://www.w3.org/2004/01/rdxh/spec#issue-tx-element


-- 
		-harry

Harry Halpin,  University of Edinburgh 
http://www.ibiblio.org/hhalpin 6B522426

Received on Thursday, 22 February 2007 16:01:37 UTC