See also: IRC log
<scribe> scribe: JK
<scribe> scribenick: JacekK
<Holger> minutes
RESOLUTION: minutes approved
<scribe> ACTION: 2006-05-09: editors of the examples doc to add a todo about updating any [DONE] [recorded in http://www.w3.org/2002/ws/sawsdl/minutes/20060516#action01]
<scribe> ACTION: 2006-05-09: EricP to make the HTML to override the schema at http://www.w3.org/2002/ws/sawsdl/spec/sawsdl [ONGOING] [recorded in http://www.w3.org/2002/ws/sawsdl/minutes/20060516#action02]
Registration for the F2F
-> http://www.w3.org/2002/09/wbs/39001/200606galway/
Holger: we need a volunteer to review the WSDL RDF mapping
<Holger> current URI http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/wsdl20/wsdl20-rdf.html?content-type=text/html;%20charset=utf-8
Holger: would we be interested in a small-tech-plenary meeting next January?
Carine: the WS groups will probably be closing at that time
... little overlap with others
<caribou> WS Desc ending Jan 31, 2007
<caribou> WS Chor ends December 2006
RESOLUTION: our group is not currently interested in a multiple-group meeting next January
Holger summarizes the issue
Holger: how did the WSDL-S people actually solve this problem (of pointing to documents with semantics)?
JMiller: is this like an import construct for us?
... would it be located in a single place like import, or close to the references?
Holger: it would be allowed close to the references
Rama: the reference is a URI so it can serve as both locator and identifier of the thing
Joel: the type of modelReference is URI so it should work
Holger: the URI of the model thing need not point to the document
Joel: can you please give actual examples on the list?
Holger: it's in our spec
<Holger> http://www.w3.org/2002/ws/sawsdl/spec/po/purchaseorder.wsdl
<Holger> sawsdl:schemaMapping="http://example.org/examples/POAddress.xsl"
<Holger> sawsdl:modelReference="http://example.org/ontologies/purchaseorder#OrderConfirmation"
Rama: the URI may actually not be helpful in getting the semantic description
<Amit> I am not able to open these urls
Rama: we don't have namespaces...
JacekK: we could just have a list of URIs without namespaces
Joel: we could have some kind of namespaces
Holger: we could just add note that the URIs should help getting the model document
JacekK: the client can have external means of getting models
RESOLUTION: we don't add modelLocation, we will summarize good practices
Holger summarizes the issue
Rama: my questions is whether we need schemaMapping at all
... other issue - why to distinguish between complex and simple types
... one proposal: only leaf-level model references, no schemaMapping
Amit: ontologies have different uses, may not necessarily work like schemas
... we may want to support both schema-level aspects and instance-level aspects
... in future we may go to RDF where a node may belong to multiple schemas
JacekK: what schema mapping are we talking about - schema-level or data-level?
Holger: I disagree with removing schemaMapping, I find it useful
... schema mapping can also go in either direction - xml->ontology or back
Rama: with an annotation I want to be able to say what an XML element means
... that's modelReference
... if the relation is not "is_a" but some other relation (conversion), we need mapping
... this is data-level mapping
... if we want both, that may be fine
Amit: even in discovery you may need data-level mapping
... even the ontologies that you'd map to may already be populated
... the mapping can have constraints on the data values
Rama: we're dealing with abstract WSDL interfaces, not with instances that have data values
... original intention of schema mapping was also to include modelReferences for leaf elements in complex type
Joel: it was about the "one-to-many" or "many-to-one" kinds of relationships between the XML and the ontology
... then there is the instance-level data mapping
... I think we combined the two and it may now be confusing
... Rama suggested we might begin by separating them completely and discussing them separately
JMiller: are these mapping as Rama proposes them (schema-level) useful for discovery?
Rama: for discovery, we don't need schema mapping
... first level of discovery is about comparing intent of my request and what the service talks about
JMiller: how does schemaMapping contribute to discovery?
Joel: we were looking at the situation where the schema has first name and last name, and the ontology has name
... for this we need more complex mapping
JMiller: with modelReferences, you'd just have a list of references?
Rama: I was suggesting that the ontology can describe what the relationship between first and last name is
... an RDF mapping would give us the relationship between the XML element and the concept in the ontology
... modelReference doesn't define the relationship
... RDF mapping can have more explicit relationships
... it can be part-of, subclass of etc.
<scribe> ACTION: JacekK to write an email trying to separate the different things in this issue [recorded in http://www.w3.org/2002/ws/sawsdl/minutes/20060516#action03]
Holger: we should talk about when we need different kinds of information
... the following issues on the agenda are related, we'll skip them now
Holger summarizes the issue
Joel: category has complex value which is different from modelReference
Holger: I understand the usefulness of category, but I don't see the distinctions between the category attributes
... we should clarify in the spec what the different attributes do
Rama: modelReference on interface helps a client make a quick determination if the interface is interesting
... categorization aims to help publishing the service in a registry
... the attributes are specific to UDDI, but it can be extensible
Joel: it's not necessarily tied to UDDI, it's in the form of UNSPC and the like, taxonomies not ontologies
Holger: I still don't understand why in category we leave the semantic world and go to a more practical model of key-value pairs
... and we could add publisher and such additional things as well
Rama: good point, we may have had publisher in a previous version
JMiller: we are trying to focus on the abstract interface level, not on concrete service properties
Joel: so Holger, you want us not to specify such things as category etc, instead we should reuse modelReferences?
... but here we are pointing to something else than ontologies, so we needed an extended construct
Holger: but we could add Publisher etc, so we should stick with a single thing, modelReference
Rama: we should think practically, what will be useful for companies with many services
... do we want to enable a user to capture categorization?
<Amit> don't think took figuring out is sufficient
Rama: yes there are other things, we should think about what will be most useful
Joel: what motivated inclusion of category is that registries only support non-ontology taxonomies
Holger: the use case document can help us clarify this
<scribe> ACTION: Holger to try and describe the use cases for category element [recorded in http://www.w3.org/2002/ws/sawsdl/minutes/20060516#action04]
meeting adjourned