WS Architecture F2F (Thursday am) -- 6 Nov 2003

Present: Mike, Roger, Zulah, Frank, David, Hugo, Jeff, Bijan_(observing), Jacek Kopecky (observing)

Regrets:

Chair: MikeC

Scribe: Shishir

<scribe_sg> Mike: Katia has volunteered her team's services to do the formal representations using OWL

Massimo's OWL Ontology

<scribe_sg> Massimo: Took the arch document & translated it into OWL. Tried to be consistent across all 3 definitions.
... Massimo: May have violated some of the rules in the resource model
... Mike: Summary of suggestions:
... Perhaps we need to be more rigorous in our language
... Frank: Some things that OWL can't capture.

<dbooth> We need to be more precise about the cardinality of relations. E.g., "X has a Y" or "X may have a Y". Exactly one? At least one? 0 or 1? 0 or more?

<Roger> Is there a processor that can make UML diagrams out of the OWL? It appears to have enough information.

<scribe_sg> Roger: How did you handle 2.3.1.4.2?
... Roger: How was "contains" handled in OWL?
... Roger: The two contains have different interpretations? the address information as against the body of the message
... The address information is part of the envelope, the body is contained in the envelope
... Bijan: A lot of machinery would be required to disambiguate that
... Mike: How easy is it to re-do this work over time as the arch document text is modified?

<Roger> Actually, I just realized that "body" is undefined in the concepts, so OWL obviously didn't handle it.

<scribe_sg> Frank: Could use the XML version of the document to generate OWL
... Roger: Content is not a concept, so OWL would not have handled it
... Bijan: Question is, is the contain relationship worth preserving? what inferences may be implied by maintaining that relationship?
... Mike: Learning point: To be normative, we need to be more precise. This exercise will bring ambiguities out
... Paul: What benefits do we have from the OWL?
... Bijan: Consistency checks, can general pictures using visual OWL editors
... Bijan: That distinction would be controversial in WSD.
... Roger: As WSA, we should point out to WSD that headers and body need distinction
... Mike: Appreciate the work done by Massimo

<scribe_sg> Mike: integration between text descriptions & formalisms?
... Massimo: If there is a mapping, OWL may help specify the english language more accurately
... Frank: Good sanity check for us & for OWL.
... Frank: Issue with UML is too much precision. Maybe it is the right time for that level of precision.
... Frank: Fully support the use of OWL. OWL tooling not there yet. OWL's big benefit is the semantics.
... Frank: OWL has a model theory. There are rules for interpreting.
... Frank: Can we use OWL within the text?
... Mike: OWL has the notion of cardinality. what about optionality?
... Frank: That's a part of cardinality.
... Massimo: Any property is optional in OWL. It may be there 0 times or a hundred times
... Frank: We could do a consistency proof for the entire architecture.
... david: We need to resolve the glitches that have already been identified
... Frank: We have to read the comments & OWL to get to these issues

<bijan> http://www.mindswap.org/2003/pellet/demo/ #Online owl consistency and sytnax checker. Code downloadable as well
... Also has some minimal visualization (of the classification)
... Feedback and feature requests welcome

<scribe_sg> Frank: Lets come back to this..
... Bijan: WSD is using RDF & OWL where possible.
... Bijan: Ownership problem for bugs & issues post WSA

<dbooth> A clarification: WSD will produce a mapping from WSDL to RDF, which Bijan reports will actually be using OWL

<scribe_sg> Bijan: If WSA wants to work with OWL, a new member could join WSA to do that work from U.Maryland
... David: That would be very useful for the sanity check part of WSA work.
... David: Not sure if it would be a part of the XML document. That's a separate issue actually maintaining the correspondance between the doc & OWL
... David: Envision little bits of OWL to be distributed throughout the document?
... Frank: The real issue is what is normative?
... David: Can't make OWL normative in this time frame. It should be a secondary & helpful exercise
... Paul: If the document is taken to REC status, then we could consider normative OWL
... Bijan: Everything from english text is not going to be captured in OWL. It would be a subset
... Frank: XML editing is hard enough for editors.
... Bijan: Massimo would have done the OWL editing in a text editor
... david: It would be an appendix to the document
... Roger: If we end in 2 months, this can't goto REC. What about the issues that are unresolved? will they fall off the edge?
... Hugo: Not necessarily. What if the W3C creates the Web Services TAG which continues to do Architecture work
... Hugo: This group would have a very issue oriented approach
... ACTION to Bijan to check if someone from U.Maryland has the resources to help WSA with OWL related work

<dbooth> ACTION: Hugo to review Massimo's OWL and resolve issues in the document that Massimo points out

<scribe_sg> ACTION: Bijan to check if someone from U.Maryland has the resources to help WSA with OWL related work

1.7 Web Service Technologies

<scribe_sg> Zulah: this text was written before we decided on SOAP/WSDL. So section should reflect that
... ACTION: Zulah to look through this section for wording that was used in absence of a decision to use SOAP/WSDL
... Zulah: Features vs. concepts
... ACTION: Frank to re-draft Concepts and decide on Features VS Concepts
... Roger: Complete VS coherent
... Roger: Complete is not appropriate.
... Roger: Use of SMTP breaks the wording in Conformance section.
... ACTION: Frank to resize meta.png

Section 2.3.5 Management Model

<scribe_sg> http://lists.w3.org/Archives/Public/www-ws-arch/2003Nov/0003.html
... Issues with existing management model: Didn't want to split management service & business service
... Roger: Does the provider provide both the service & the management service?
... Zulah: You just have to aggregate a management interface over the existing interface
... Roger: Issue with representation of Manageability interface as the managed functional interface is not called out explicitly
... Roger: Interfaces are recursive. Is that true in WSDL?
... Frank: No.
... Frank: Our use of service is not consistent with WSDL
... Roger: I would expect to get one WSDL from a URL that has both the functional & manageability interfaces
... David: WSDL 2.0 allows interface inheritance
... Zulah: Manageability of a service is a property of the service.
... Zulah: Can you bind certain interfaces with certain endpoints and others with other endpoints?
... Frank: Can the manageability property be projected out of the managed service out as an individual service?
... Zulah: The WSDLs could be independent or the same.
... Zulah: Are we architecting for WSDL 2.0?
... Frank: Yes
... Jacek: in WSDL, when we discuss targetResource, management comes up often.
... Frank: Different services have different URIs
... David: Depends on what your notion of a service is. If you use separate WSDLs for func & mgmt, the service parts of those two WSDLs are independent
... David: targetResource is being considered to identify the potentially common real-world service
... David: WSDL does not give you anything to show the relationship between two interfaces.
... David: Other mechanisms exist for that.
... Zulah: It will be very hard to flatten out the model & express what the management group is trying to express.
... Frank: This goes beyond management. It is not possible to have ONE WSDL file to exist.
... David: Both a management WSDL and functional WSDL can be true & applicable at the same time
... Frank: They can't have the same QNames.
... Frank: Its really important for a service to have one interface, conceptually
... Frank: WSDL 2 will have some real-world mechanisms to model this
... David: Think they can have the same QName (so they point to the same URI)
... Mike: How do we resolve this?
... Frank: Question for WSDL: Can you cope with projection of the service across multiple WSDLs?
... Roger: What about different bindings?
... David: You use different WSDLs for different viewpoints
... WSDL does not have to have a URL, but does have a targetNamespace
... Roger: What about discovery?
... When you have a WSDL, the meta data related to manageability, etc is discoverable
... Zulah: Issue of Cardinality
... Zulah: interface generalizes the manageability interface. how does aggregation distinguish from generalization in the diagram?
... Frank: Need to normalize the terminology between WSA & WSD regarding WSDL & service
... Zulah: A WSDL service should actually be a service view
... A description would be only applicable to service views. This would be a good resolution for both WSA & WSD groups
... A WSDL service has one interface. Each description would describe one service view.
... Frank: Second question: Can a master WSDL be constructed?
... Roger: As long as they are logically related, it could be simply a list of links
... David: Inheritance can be used to represent hierarchy
... David: WSDL will probably respond that there's nothing in WSDL 2 to prevent this model
... Jacek: There's a cacheability issue with using the same QName
... Jacek: Caching - If I have a URI, I re-use it for the defined time.
... For a Namespace, then the assumption is that the properties defined are not described differently elsewhere
... Hugo: This sounds like content negotiation.
... Hugo: This is resolved using media-types
... The debate has been discussed & rejected resolutions by targetResource (reason: ), name (reason: because of Qname caching)

<hugo> ... the differerence here is that we are doing content negotiation _within a WSDL document_
... [ completing what I said ]

<scribe_sg> ACTION: DAvid, Zulah to create text to communicate with MTF regarding this

<dbooth> ACTION: dbooth and zulah to compose a message to the WSD WG asking if two WSDL documents can reference the same service and therefore provide different views (or projections) of that same service

<scribe_sg> Frank: The new diagram's obligation model is closer to the Social model rather than the Policy model
... Zulah: The Management Policy needs to be independent of the Policy Model
... Frank: Can use a contract/agreement to define the acceptance of obligation
... Frank: Provider ensures/has a management policy
... Zulah: May add policy to the service model along with Actions
... ACTION: Frank to resolve policy for the service model

3.1 Web integration

<scribe_sg> 3.1 removed from doc
... 3.2 removed from doc
... 3.3 removed from doc
... 3.4

<scribe_sg> Frank: Should add the using a WS stuff in here
... Frank: A stakeholder is any entity with interest in WS (at a global level)
... Mike: Its more about perspective..
... Frank: refered to as *ilities previously
... Zulah: What does this section 3 accomplish?
... Mike: We don't have the time or resources to work at the level of detail that's perhaps required in this section
... We need to reword the intro text

<mitrepauld> roger... stakeholder section now holds things that satisfy our requirements, not satisfied by section 2

<scribe_sg> Mike: Lets have more topical discussion as against meta-discussion

<scribe_sg> Discussing the discovery process
... 4 step process: becoming known parties; agree on SD + semantics; etc.
... Paul: talking about runtime VS design time helps in discovery discussion
... Paul: 3.4.2 could have discussion of runtime VS design time
... Roger: think of 4 quadrants around a design-runtime axis and manual-automatic axis
... 3.4.4: We need to talk about federation of registries
... discussion about importance of federation in WSA document
... ACTION: PaulD to propose text on federation of registries
... 3.4.3: We need to have consistent use of Trust
... Is it trust or control? should it be here or in the security section?

<dbooth> ACTION: Zulah to give dbooth her suggestions regarding "Trust and Discovery"

<scribe_sg> ACTION: David will find new home for 3.4.3 Trust and Discovery & point to it from Discovery
... Zulah: Does the re-named section adequately cover using WS over and beyond discovery?

3.5

<scribe_sg> Frank: Not sure if its a good idea to have Semantics as a stakeholder
... Better in "How do you make sense of WS" section

<mitrepauld> perhaps section 3 should take some use cases and point to things in section 2 that support them.
... I have to drop off the phone now.

<scribe_sg> Frank: looking to rewrite 3.5 with things like visibility, ...
... because it is a part of the semantics of the services, and the standardization of the messaging
... Discussion of Metadata, semantics & policies
... Frank: Semantics of SOAP intermediaries can be understood as an architectural way to interpret semantics
... ACTION: Frank to propose text around architectural approach to semantics (intermediary visibility issue)
... Frank: OWL is the language to write the formula. UML has no place to define the meaning of relationships similarly
... ACTION: Frank to see if deleted text about syntax & semantics in introduction should move to stakeholders semantics section

3.6 WS security

<scribe_sg> Zulah: Distribute Policy across multiple sections beyond Security
... Roger: For example, EDI does not have buffer overruns, but that kind of architecture comes with its pros & cons
... Roger: Other policies represent other threats, and we need to talk about it
... Frank: Security breaches occur because of breaking abstractions. Not sure if this is the level to address such breaches
... Roger: Nail down security threats and offer guidance are things we are expected to do
... Mike: Need a security expert to do this work
... Roger: GXA describes such security threats really well
... ACTION: Roger to contact Abbie and see if he will work on an enumerated list of security threats

3.7

<scribe_sg> 3.7, 3.8, 3.9 deleted
... Zulah: A pub-sub event model is missing from this document.
... 3.11 deleted
... ACTION: Mike to add and wordsmith text in 3.11 choreography
... David: Looks reasonable already
... Mike: Will add prespective

<scribe_sg> ACTION: Mike to check archives to find text we think we agreed to for message reliability & add it to the document

<scribe_sg> discussion on recovering lost reliability text
... ACTION: Mike to propose changes to WS Reliability section in stakeholders perspective
... Frank: Some of it can go into the concepts section
... ACTION: Roger to send feedback on overall reliability

3.14 WS Manageability

<scribe_sg> discussion on using UML diagrams
... Mike: For document consistency, there's some discontent around levels of details in various sections
... ACTION: Zulah to propose a new WS manageability section in stakeholders section

3.15 Web services technologies

<scribe_sg> David: Don't need this section at all
... David: Lets replace this with a list of architectural issues?
... Mike: Proposed outline would be XML, SOAP, WSDL & what we've done with the list of lists with pointers
... Also, perhaps mention the W3C Chor work as its a W3C group
... Roger: doing that and not mentioning the OASIS WGs will reduce credibility of this document
... David: We point to latest message on the list & credit Roger & Paul Denning
... Mike: Sure as long as its not normative
... ACTION: Roger to go through & ensure he's OK with the latest list of standards
... ACTION: Editors to reference the list of standards from 3.15 WS technologies in the stakeholders section
... Discussion of section 3.15 Web services technologies and 1.7 Web services technologies

3.15.1 XML & Web services

<scribe_sg> Frank: Do we need to discuss XML Schema?
... David: WSDL is based on XML Schema
... Mike: Talk about dependancies of SOAP and WSDL on various XML technologies

3.15.2 SOAP

<scribe_sg> Mike: current text does not correlate to the concepts defined in the document.
... several word-smithing items and specification versions added into document

3.15.3 WSDL

<scribe_sg> Mike: Is it important to capture "what does a WSDL document describe"?
... David: Already done in servuce description section
... so this WSDL section should reference that service description section 1.5.3
... Mike: Why don't we define binding? Because it is implementation detail.
... Acknowledgements discussion
... Need to apply good & bad standing rules
... Roger: This group should be able to speak to the documents produced by the WG
... David: Will be happy to help with the clean-up process

<dbooth> ACTION: DBooth to provide MikeC a list of WG members in good/bad standing
... ACTION: Hugo to provide boiler-plate message verbiage about WG members who have been absent and will be removed from the WG membership list unless they take action

WS Architecture F2F (Thursday am and pm)
6 Nov 2003

Attendees

Contents

Massimo's OWL Ontology

1.7 Web Service Technologies

Section 2.3.5 Management Model

3.1 Web integration

3.5

3.6 WS security

3.7

3.14 WS Manageability

3.15 Web services technologies

3.15.1 XML & Web services

3.15.2 SOAP

3.15.3 WSDL

Summary of Action Items

WS Architecture F2F (Thursday am and pm) 6 Nov 2003

Attendees

Contents

Massimo's OWL Ontology

1.7 Web Service Technologies

Section 2.3.5 Management Model

3.1 Web integration

3.5

3.6 WS security

3.7

3.14 WS Manageability

3.15 Web services technologies

3.15.1 XML & Web services

3.15.2 SOAP

3.15.3 WSDL

Summary of Action Items

WS Architecture F2F (Thursday am and pm)
6 Nov 2003