Re: SAWSDL Spec: Some Comments


We have discussed the issues raised by some folks at IBM during SAWSDL spec
review (as part of the' Last Call ' review of SALWSD spec) at the SAWSDL
Working Group meeting last week again and I would like to add the following
clarifications to the responses that I already provided in the e-mail

Rama Akkiraju wrote on 11/01/2006 04:31:13 PM:

> All,
> I reviewed SAWSDL spec with a few interested folks in IBM (mostly IBM
> Research) and have received the following comments. I'm typing them up
> on their behalf along with the responses that I have already provided. I
> wanted to bring it to the attention of larger team for additional
> 1. Examples for associating Preconditions and Effects with Web Services:
> Many people in the audience felt that they would be using SAWSDL
> modelReferences for representing preconditions and effects associated
> Web services and that they would find the spec and usage guide incomplete
> without at least a brief discussion on preconditions and effects and a
> examples to show how to do it (sort of like how we show examples for
> publishing Web services in registries)
> (My response:  We were explicitly asked to not 'discuss' preconditions
> effects in SAWSDL Working Group as part of our charter and so we didn't.
> Personally, I prefer that we provide some examples for associating
> preconditions and effects with WSDL documents in the usage guide. We
> discuss this in the Working Group)

Request to add examples for preconditions and effects have been received
from multiple sources and I'm glad to report that we have agreed that we
should provide examples for associating preconditions and effects using
'modelReference' extensibility attribute in SAWSDL Usage Guide. We will be
addressing this in the coming few weeks. Please feel free to contribute any
examples that you may already have based on your work.

> 2. When we associate multiple annotations with an element using
> 'modelReference' (eg. below) people felt that not providing an explicit
> relationship (such as 'intersection') could lead to arbitrary
> interpretations.
> (eg:
> <xs:simpleType name="itemCode"
> sawsdl:modelReference="SampleOntology#PartNumber SampleOntology#SKU"/>
>     <xs:restriction base="xs:string"/>
>   </xs:simpleType>
> )
> (My response: We defer to the ontology(ies) to provide relationships
> between multiple annotations as explicating any relationship in a WSDL
> document at the modelReference level could create inconsistencies and
> multiple places for specifying such relationships)

My response above holds good. Nothing more to add on this point at this
time. If some one still has concerns about this, please raise the issue
with specific examples. Or if you have any concrete change suggestions
please send it along.

> 3. modelReference Vs. lifting and lowering schema mapping extensibility
> attributes: Some people felt that they did not appreciate the distinction
> between modelReference and lifting and lowering schema mapping
> extensibility attributes since all three were meant to serve as URI
> pointers. The question was why not just have one extensibility attribute
> called modelReference and use it for everything.
> (My response: modelReference is meant for associating semantic-level
> annotations while lifting and lowering schema mappings are meant for
> associating mappings to transform data values associated with schemas or
> ontology instances. Although all three use URIs as mechanisms for
> this, we feel that their purposes are different enough to warrant
> extensiblity attributes)

A couple of additional clarifications to add to the point on why
'modelReference' and 'liftingSchemaMapping''loweringSchemaMapping' are not
the same concepts.

1. schemaMappings are not direct pointers to concepts in a semantic model
as modelReferences are. Schema mappings are transformations that are
specified in a transformation language of choice. These transformations may
point to the semantic data in an ontology but the mappings themselves are
not direct pointers to concepts in semantic models as modelReferences are.

2. When multiple URIs are specified on liftingSchemaMapping or
loweringSchemaMapping, the schema mappings they reference are to be treated
as alternatives, i.e. the client processor should choose one of them to
apply, and the choice is fully at the client processor's discretion. For
example, a mapping can be selected based on what mapping language the
processor supports (different alternatives can use different languages),
based on the availability of the mapping document, or by other preferences.
We restricted the interpretation of multiple URIs to be alternatives in the
case of schema mapping because arriving at multiple XMLs does not help when
you are trying to invoke a service. Whereas the interpretation for multiple
URIs on a modelReference is that they 'all apply'. One of the resons is
that one might want to annotate an XML element with multiple semantic
concepts to enable matching during discovery.For example, I may want to say
that a 'name' element in my XML schema points to both 'GivenName' and
'FirstName' concepts in an ontology to enable matching with multiple Web

I hope these points illustrate the difference between 'modelReference' and
'liftingSchemaMapping' and 'loweringSchemaMapping' notions.

> Regards
> Rama Akkiraju

Received on Tuesday, 21 November 2006 22:23:24 UTC