Minutes, 16 Sept 2004 WS Desc FTF

Web Services Description F2F
Thursday 16 Sep 2004
See also: IRC log [http://www.w3.org/2004/09/16-ws-desc-irc

 David Booth            W3C
 Helen Chen             Agfa-Gevaert N. V.
 Roberto Chinnici       Sun Microsystems
 Glen Daniels           Sonic Software
 Paul Downey            British Telecommunications
 Hugo Haas              W3C
 Hao He                 Thomsona 
 Tom Jordahl            Macromedia
 Anish Karmarkar        Oracle
 Kevin Canyang Liu      SAP
 Jonathan Marsh         Chair (Microsoft)
 David Orchard          BEA Systems
 Bijan Parsia           University of Maryland MIND Lab
 Arthur Ryman           IBM

 Allen Brookes          Rogue Wave Software
 Youenn Fablet          Canon
 Jean-Jacques Moreau    Canon

 Amelia Lewis           TIBCO
 Asir Vedamuthu         webMethods 
 Sanjiva Weerawarana    IBM

Thursday 16 September
[dbooth: Scribe: PaulD]

09:00  Primer suggestions from Hao

Hao:      How dynamic is the WSDL, can F&Ps be used to describe 
          current status of the service etc?  Is WSDL only for use 
          before deployment time?
Jonathan: It's just meta-data
Hao:      Do we need to talk about use of WSDL in discovery?
Glen:     No, though we do talk about comparing WSDL contracts useful 
          in discovery.
Hao:      Wonders where policies which are dynamic will be published.
Paul:     Suggests Hao looks into the C&C workshop will cover these 
          very issues.
[GlenD:   Constraints and capabilities workshop -   
Hao:      Given an endpoint, how do i find the WSDL describing that 
[Rehash of the whole "what lives at the end of a namespace URI"

09:30 Topic: RDF

Slide 2 URIs
Roberto:   If the mapping is normative on the component model will this 
           add requirements to the processor.
Bijan:     No, it's a separate module which could be implemented 
Jonathan:  Component URIs could have other uses
           for Bijan's slides]
Roberto:   Doesn't the use of a frag-id + wsdl media type force this on 
           all conformant processors?
Bijan:     You can write a processor that doesn't implement or use
           it's implementation dependent.
Discussion of packaging of work in Part 1 or as a separate document
Jonathan:  Proposal is not to move component designators but make them 
           normative due to dependency upon media type registration
[Roberto:  And move them nearer to appendix A (media type registration)]
Arthur:    Fragment syntax should be normative and part of spec.  Use of

           component designators by RDF should be kept separate
Jonathan:  Outlines reordering of part 1 on the screen: frag-id part
           normative allows frag-id and media type to reference each
RESOLUTION: Move App C to RDF Mapping spec, except the frag-id which
           move within media-type reg appendix.
ACTION:    Editors to move App C to RDF Mapping spec, except the frag-id
           which will move within media-type reg appendix.

Bijan:     Component type and property URIs
Bijan:     When constructing URIs will use camelCase for consistency
RESOLUTION: Camel case is OK for RDF mapping constructed URIs.

Bijan:     Definitions, targetNamespace
Arthur:    Can we change top level element from 'definitions' to 
[The crowd goes wild at Arthur's excellent suggestion, but pushes it on
the stack for now.]

Discussion of relationship between component designators and the
Roberto:   We could get rid of the top level component, Schema doesn't
Bijan:     Top level component does provide some scoping which is useful
           some RDF applications
Glen:      Same WSDL document could be at many different locations,
           choice to have the top level URI inside the document.  Why do
           want the definitions component at all?
Arthur:    Provides scope: what is the set of components i'm talking
Glen:      If you're just including a bag of other components possibly 
           from elsewhere, why would identifying the container itself be

Arthur:    Set of components likely to be unique for each URL, it's 
           accidental if two separate documents end up having the same 
           set of components.
Bijan:     Can you have extensions to the top level definitions?
Roberto:   Yes, e.g. choreography 
Jonathan:  No change to the spec!
Bijan:     Now has better understanding of status quo and has no need 
           for top level namespace in his mapping
Jonathan:  targetNamespace could be moved down into "XML Representation"

           to reduce its preceived importance
Paragraph 6-9 of section 2.1.1 moved into 2.1.2 which talks about the
[dbooth:   Also note that this sentence is wrong: "The components
           defined within a single Definitions component are said to
           to the same target namespace."]
ACTION:    Editors to fix paragraph 6-9 of section 2.1.1 moved into
           which talks about the syntax.
ACTION:    Arthur to write up his proposal to rename 'defintions' to

Jonathan: let's take a break before getting into this ..

10:30 Break     ----------------------------------------

11:00 RDF questions (cont.)

Bijan:     {message exchange pattern}
[camelCase seems OK with everyone.]

Bijan:     MEP - is it a URI or 'individual'?
Bijan:     Wants to deviate from the spec by using an as a string value
           rather than a URI.
Tom:       Is the RDF mapping both ways?
Bijan:     No round-tripping, in particular there is loss of imports.
           Use of mapping is not as a model, but for 'saying more 
           specific things'.
[Seems OK with everyone.]

Bijan:     Interlude
Bijan:     RDF Extension
Bijan:     Wants a rdf:RDF property everywhere there is an extensibility

[Seems OK - we have an open content model, this is allowed.]

Jonathan:  Progress on RDF spec?
Bijan:     Should have something in the next couple of weeks.

11:30 Administrative items

Jonathan:  Meeting in January - possibly Sri Lanka or Sydney?
Jonathan:  Topics will include moving to CR in particular testing, the 
           test suite and implementations
Jonathan:  Tech plenary end of Feb/March in Boston
[No decision, we'll see where, and whether we even need, a Jan meeting.]

Paul:      Any progress on IP for test suite contributions
Hugo:      Team has been working on this issue - expect something posted

           to the AC list soon. Anticipate that test cases will be 
           published under the document license
[hugo: Document license:
[hugo: Software license:
Jonathan:  You are free to contribute using these licenses regardless of

           a W3C policy for a test suite.

Jonathan:  How do we expect a bake off to work - expect to offer an 
           open invitation.  Should be publishing expected test cases 
           and bake off scenarios by November.
Glen:      Need to have different classes of tests.  Suggests getting
           involvement of Bob Cunnings of SOAPBuilders and WhiteMesa in 
           the planning
ACTION:    Glen to contact Bob Cunnings regarding planning for CR

11:45 Last Call issues [9]
    - (In issues list order, with Asir's last if possible)

 [9] http://www.w3.org/2002/ws/desc/4/lc-issues/

Topic: LC32
[hugo: s/32/31/
[tomj: I believe we need at least the statement that an XML 1.0 WSDL
must validate with our schema for conformance
can our schema for XML 1.0 serialisation also be applied to an XML 1.1
ACTION: editors to remove text regarding schema validity from section
1.2 and add text to describe schema is for the conformance of XML 1.0
serialisation and the intent is it may be applied to other future
serialisations such as XML 1.1
discussion: description of contents of properties throughout part 1 uses
schema 1.0 types. this could preculde moving to XML 1.1 serialisations
if types such as xs:string change
Paul: propose we use our own types wsdl:string etc and avoid using
schema language when describing allowed property values
Roberto: thinks infoset should resolve this issue
Roberto: alternative proposal, rename these sections as XML 1.0 specific
[Roberto: http://www.w3.org/TR/xml-infoset/#infoitem.attribute
ACTION: editors to make clear that mapping to XML is 1.0 specific
Jonathan: break for lunch

12:00 Lunch     ----------------------------------------

Scribe: Tom Jordahl

13:00 Last Call issues (cont.)

Issue LC34a: Completing Part 1 Appendix C: URI References for WSDL

Hugo:     Appendix C is very useful. In lots of places we note that 
          things cannot be referred to by QNames, so we should point 
          down to Appendix C.  For instance in section 2.3.1 we could 
          add a sentence at the end of the paragraph that talk about
Jonathan: Proposal - in places where we talk about how to access
          which can't be accessed by QName, add text pointing to
Appendix C.
          We should pull the section that we moved out of C (C.1) and in
          the RDF (earlier today) and put it back.
Roberto:  Objects to that. Likes it in RDF spec
Bijan:    I don't care if we move it back
Hugo:     Supports Jonathans proposals
Consensus - Add links part of Jonathans proposal.
Jonathan: where do those links point? 
Hugo:     Its OK to point to the places where we moved Appendix C in 
          the morning.
RESOLUTION: Add links to sections of specification where we note that 
          QName is insufficient to identify a component.
ACTION:   Editors to add links to component designators to sections
          Where we note QName is insufficient.

Issue LC34b: Completing Part 1 Appendix C: URI References for WSDL

Proposal:  As stated in the issue PLUS feature and property additions
Arthur:    Properties could have the same path and lead to 2 different 
Discussion about how the fragment identifiers work - How/if we can apply
them to features and properties
Arthur:    Maybe we should assign a task to the author of Appendix C to
           features and properties?
Discussion on how adding a feature/property syntax in existing selectors
might work
Proposal:  Accept resolution in issue, Add action to do F&P
Arthur:    - but scrub the syntax to be more concise
Arthur:    Volunteers to scrub syntax to make sure everything makes
RESOLUTION: Close issue 34b - accept proposal in general, include F&P
            additions too
ACTION:     Arthur will scrub the proposed syntax and create Features
            properties syntax.

Discussion about features and if they have to have a unique URI
Glen:      two F/P with the same URI in a scope should be an error.
    <feature uri="uri1" />
    <feature uri="uri1" />

Proposal:  Specify that the above is an error
Jonathan:  This will help the fragment ids work correctly
ACTION:    Editors to implement the restriction that feature and
           URIs must be unique in a scope.
Anish:     Do scoping rules cover the case where the same uri is used in

           nested scopes?
Glen:      Yes, scoping rules apply - the nearest one wins.  Raises 
           possible issue on changing properties to specify that
           properties must be 'compatible' with previous specified
Some discussion about why this would be needed and how it would work. 

Topic: LC34c: Completing Part 1 Appendix C: URI References for WSDL

RESOLUTION: Issue 34c Accept the proposal - change names to match
            names in table C-1.
ACTION: Arthur to revamp table C-1 with component names

Topic: LC34d: Completing Part 1 Appendix C: URI References for WSDL

RESOLUTION: Table will be reworked to be more readable when new 
            components are added.
ACTION: App C editors to rework table to be more readable when new
        components are added.

Topic: LC33: Part 3 SOAP Binding: default HTTP method 
Tomj:     Having trouble understanding what the problem is
Hugo:     We are talking about an extension that doesn't have syntax to 
          change its value.  We should remove the default from "default 
          value of the {http method} property"
RESOLUTION: close issue 33 - Remove default from text in bullet item 
            "HTTP Method Selection" 
ACTION: Editors to remove 'default' from text in bullet item "HTTP

Topic: LC37: Part 3 3.6.4 Mapping Between HTTP Operation's XML
Representation to Component Properties and default values 

RESOLUTION: Accommodate the method default and input serialization 
            default in part 3, section 3.6.4
ACTION: Editors to redo Part 3 Section 3.6.4 to accommodate method
        and input serialization defaults.

Topic: LC38: Part 1: DTD as the schema language for WSDL 

Tomj:     Why is this in the spec?
Jonathan: Jacek probably wrote it. Shows our ability to handle other 
          type systems
Tomj:     Not inclined to put a 'real' URI there - lends it weight 
          where there isn't any
Dbooth:   What is the cost? Nothing, why not do it.
Jonathan: We should do something as it is specified with more rigor than
          example.org URI indicates
Jonathan: Should we remove this?
Roberto:  Type system extensibility is a pretty big deal, these examples
          really help.
Jonathan: Gives the impression of possible interop, but we really don't 
          know if it would work
Jonathan: we have the Relax NG, lets drop DTD
[pauld:   DTD type system mentioned in this article!
Hugo:     Put real URI there, flag it in CR stage, remove it if nobody
Proposal: Remove appendix D
OR define a real namespace URI
Straw poll.
 Option 1 - remove Appendix E.1
 Option 2 - Assign a real namespace URI
 Option 3 - Leave status quo (example.org namespace)
 Option 1 - 0
 Option 2 - 8
 Option 3 - 2
 [jjm: option 3]
RESOLUTION: close issue 38 - Assign a new namespace URI.
ACTION: Hugo to get a URI to use for DTD example in Appendix E.1

Topic: LC12: "whttp:location" attribute is missing 

RESOLUTION: close issue 12 - Add whttp:location to section 2.1, the 
            binding operation element (same place as 
ACTION: Editors to add whttp:location to section 2.1, the binding 
        operation element (same place as whttp:defaultTransferCoding)

Topic: LC13: HTTP Operation Component

Jonathan:  Appears to be editorial?
Roberto:   Should be HTTP Operation Extension
RESOLUTION: close issue 13 - change use of HTTP Operation Component
ACTION:    Editors to change use of HTTP Operation Component to
           like Binding Operation Component augmented by HTTP

15:00 Break     ----------------------------------------

15:20 Last Call issues (cont.)

Topic: LC14: Mapping ref attribute to {fault reference} - Type Mismatch 

Jonathan:  Change the mapping that the value is indeed the component. 
           The mapping should be corrected
RESOLUTION: close issue 14 - Fix the mapping to say the value is the
ACTION:    editors fix the places described in issue 14 to say the 
           value is the component.

Topic: LC16: Interface = design of the application

RESOLUTION: close issue 16 - accept the proposal to remove the 
            sentence: "Thus, an interface defines the design of 
            the application."
ACTION: editors to remove the sentence "Thus, an interface defines 
        the design of the application" in section 2.2.1

Topic: LC17: URI Serialization: Order may be Lost

Jonathan:  Schema lists have order
Tomj:      How can we fix this so that the URL parameters don't lose 
           that order?
Arthur:    You can get the full query string if you care about order
Jonanthan: Doing nothing seems to be the thing to do, at least till Asir

           is present
Tomj:      We should specify in the spec that the order of the query 
           parameters is in the order of the list.
Propose action:    Editors to clarify section to state that
           list values, the order of the values in the list are 
           preserved in the url parameters.
Jonathan:  Should the order count for all the other elements?
Tomj:      Yes, they should be in document order
Jonathan:  Would prefer that the proposed resolution be delayed till we 
           have the whole solution.
ACTION:    Jonathan to record proposed resolution in the issues list

Topic: LC18: Relationship between Features and SOAP Modules 

skip for now - covered by Glen's mail to someone else, probably

ACTION: Glen to CC Asir on mail to Marc re: SOAP modules and features

16:00 Adjourn 

Received on Saturday, 18 September 2004 00:50:27 UTC