Web Services Architecture Face to Face Meeting -- 14 May 2003

Present: Chris Ferris (remote) Dave Hollander, Dave Orchard, Doug Bunting, Kevin Liu, Martin Chapman, Mark Jones (remote), Mike Champion, Roger Cutler, Davie Booth (remote), Eric Newcomer (remote), Hugo Haas, Jeff Mishinsky, Paul Denning (remote), Ugo Corda (remote) Yinleng Husband

Regrets:

Chair: Mike Champion

Scribe: various people

<DaveH> good morning...
... how about getting to work
... hi
... topic 1 - sync and async
... where do we need definition?
... daveH volunteers to scribe
... Meeting Objectives - advance the draft as far as possible!

<mchampion> We need to determine what it means to produce a "successful" WSA document.
... Then we need to do thru the document to determine where each stands in terms of the success criteria:
... - Things missing from the document that are necessary for success
... - Things in the document that are more or less successful (net wordsmithing)
... - Things in the document that are needed for success but need lots of work
... - Things in the document that are NOT necessary for success and can be jettisoned
... To illustrate, Dave talks about synch/asynch and MEPS, which was being informally discussed before we convened
... Section 2.2.22 on MEPs ... issue is whether we need to rigorously define this? Does the synch/asynch definition need to be made more rigorous?
... Roger: synch/asynch discussed in usage scenarios
... roger: personally thinks that our definition is "low quality" BUT a more rigorous definition is not REALLY needed to acheive WSA success

<mchampion> YinLeng: What's the relationship of our glossary to the larger W3C effort?
... Hugo: Unclear ...
... Roger: "In view of the informal way that the terms are used, it is recommended that documents should not rely upon the use of "synchronous" or "asynchronous" in any normative specification. " this is a big bogus ...
... DIscussion: is this "major rework" or "wordsmith"?
... Roger: rework
... roger: but not high priority
... dave: low prio wordsmith ... Hugo and YinLeng and Mike agree.
... ACTION: add "try to improve wording of glossary entry for synchronous" to issues list, low priority

<mchampion> Dave: how about MEPS definition itself.
... Mike: I think we need to clarify MEP vis a vis WSDL and choreography
... Hugo: Interesting text from Mark Jones authored in Arizona, we need to revisit.

<mchampion> Dave: ed note in section 3.12 (conversations).
... Informal poll: Chair thinks that consensus is that MEP definition must be improved, and it's high priority.
... Roger -- we need to "drill into subject" not redefine it.
... Dave: This will help us clean up a lot of text.
... No dissent.
... ACTION: Chairs will put MEP drilldown discussions on agenda(s) in near future.
... Moving into "how do we define success" discussion.
... Roger: Most fundamental expectation is that we define "web service" and how they fit into world
... Dave: agrees, but this seems to be something WG can't agree on.
... Dave: We can best get consensus by decomposing definition into specific properties.
... Roger: agrees with approach.
... Dave: for example, "interface" is not considered a concept. Others agree
... Dave: Does anyone think that doc now "forms a coherent picture of web service" adequately?
... Apparently no one does think we have succeeded yet.
... Yinleng: Before we define architecture, we have to define "web serivce"
... mike: take blame ... thought that a year ago when we first started into this trout pond
... Clarification: "before" does not mean chronologically, but it means a logical dependence
... Roger: need to do this iteratively ... scope, definition, scope, definition ...
... Dave: a definition of WS should *be* the concepts, constraints, and conformance statement
... Proposed statement of consensus: It is NECESSARY, but not sufficient condition of success that we define "WSA Web Service" by reference to the concepts, relationships, contraints it exhibits.

<DaveH> This approach is a change in direction deemed necessary by the chairs to make sufficient progress. The change may only be in the context of the conversation, not the document effort

<mchampion> Hugo, Dave: Editors made this change awhile ago, we're just making it clear that starting from the "what is a web service" is pointless.
... Dave: if we start at section 2 and keep track of status, preconditions, and technical details
... Roger: notes recursive definitions in document ... "discovery service provides service to discover ..."
... Hugo: this may identify concepts that are too vague to be in doc
... Dave: we need to ground definitions in fundamental concepts
... Roger: it may be beyond the wit of man to do much better than this definition of "discovery"
... http://members.tripod.com/TheoLarch/turtle.html "It's turtles all the way down"
... Dave: Let's just start at section 2 and iteratively define "success" just like we are iteratively defining "web service"
... Dave: TimBL cornered me last week and said "what I want is that big picture" that defines what all this is about and how it fits together.
... Roger: I had a similar conversation with TimBL
... Break time ...
... DaveH: With any luck the turtles will eat the trout
... Conclusion: We will adopt TimBL's informal definition of "success" -- we will succeed when we can define Web Service and show how the pieces (description, messaging, choreography, etc. etc. etc.) fit together.

<DaveH> Para 1 - add "are defined" and change interopability to architecture
... normative - we do not have a conformance statement, therefore can not be used here.
... Action: para 2 - and concept that summary and relations are more definitive/precise and description is for general understanding
... ACTION: section 2.1 - Action: para 2 - and concept that summary and relations are more definitive/precise and description is for general understanding
... section 2.1 - good phrase: why agents are important to the archicture.
... Agreement in alphabetical order

<mchampion> Discussion of glossary vs "concepts" section
... No consensus ...
... Discussion of paragraphs in section 2.1 that use the term "conformance"
... Straw poll: who thinks that we ARE trying to define a definitive architecture against which "conformance" is meaningful
... (all but 2 raise hands)
... YinLeng: asks for clarification
... DaveO: we're talkiing about an abstract level of conformance, not something that would guarantee interoperability, and is specific to sections not global to whle document.

<mchampion> Hi Chris ... we don't have a phone in this room so you can't call in but we'll try harder to minute discussion better.

<mchampion> Proposal: these paragraphs should be marked for major rework: 1) defined in another section (at editor discretion); 2) document will identify specific concepts that would define conformance;
... 3) Each passage thaat is "normative" will be marked as such
... Martin: we also have to clarify what "realization of the architecture" means.
... DaveH: Implementation of the architecture must also be defined; is a spec or a product an "implementation"
... Chris, any thoughts?

<DaveH> Normative = impacts conformance --- we may or may not use the term normative
... definitive vs rational

<mchampion> Martin: distinction between definition and rationale... definitions become normative, rationale does not.

<DaveH> vote on ammended proposal:
... "defined in another section" means that there may need to be separate section that decribes conformance and that section will be referenced here

<mchampion> Mike: My only objection to all this is the practicality of getting consensus on this in the current environment, not on the desirability of these objectives

<DaveH> ACTION: rework two paras in section 2.1 that address conformance as described in the above proposal

<mchampion> ACTION: Editors will put in a paragraph in 2.1 making the point that "concepts" are nouns and "relationships" are verbs
... We realize that relationships are literally nouns ..."results in" == "result"
... DaveH: Some of the relationships definitions are not well grounded. Do we like the form of the "agent" part or the "correlation" part in the current doc?

<mchampion> DaveO: Correlation is really really hard to define
... We're comparing the current draft of the "agent" and "correlation" sections. DaveH feels strongly that "agent" is a good prototype; "corrlation" is full of imprecise, ungrounded concepts.
... Roger (and Mike) agree.
... Agreement that "agent" is a good template.
... Hugo: 2 ways to get "correlation" into proper style 1) define all the words so that nouns are nouns. (remember the turtles, of course)
... 2) Simplify sentences into "triples"
... <grin>
... Going to lunch, and after lunch we have a joint session with WSD (probably minuted on their IRC channel)
... Next topic: what is the top level of the "spaghetti" diagram.
... DaveH: Bottom level is the "turtles" -- foundation terms that are used unadorned
... The break is over at 1:30 CET (7:30 EDT)
... We are on break
... We are reconvening in the WSDL room ... I'll note anything important to WSA in this channel.
... Yes we will be available by phone at 4:00 EDT
... Anyone who wants to listen in before then should ping us on IRC

<mchampion> Three items for joint session: future meetings, single interface and "what is a web service", OWL demo/discussion
... Sept F2F, Description will go first
... Proposing week of November 3 for subsequent F2F, overlaps with the Query/XPath/XSLT week
... Needs to be confirmed with proposed host, Fujitsu

<mchampion> ACTION: Confirm dates with hosts
... Some discussion of having January F2F in Australia if we still have a hosting offer on the table ...
... Discussion of service / endpoint / interface
... WSDL puts a scribbled drawing on the table, may be in their minutes
... In WSDL, a service was a collection of endpoints., no rules of how they were associated. No requirements to point to same "service" or implement same interfaces
... In WSDL 1.2, a service must implement the same interface, and the "identity" of the underlying resource must be the same
... Use the term "resource" because it is analogous to Webarch concept of a resource.
... Actually, this is in section 2 of the current WSDL draft
... So, there is a single interface associated with a service, and the endpoints in a service are semantically equivalent
... Martin: the key reason to talk about abstract "resources" is that one instance of a resource maintains common state.
... By putting different interfaces in the same service element, you are asserting that they refer to the same resource.
... Dave Orchard clarifies that all these things on the WSDL diagram with URIs are "resources"
... Flopping around in the usual trout pond about resource identity, URI comparison, and the many different senses of the word "service"
... "Collection of WSDL:services" does not have a name yet.
... This model has consensus of WSDL WG, but editors have an action item to incorporate it into document.,
... One resource can have multiple interfaces, e.g. a "regular" interface and a "management" interface.
... WSDL offers an optional way to define the resource URI so that one can make assertions about the equivalence of different interfaces
... Hi Mark ... we're in the joint WSA - WSD meeting talking about "what is a web service"
... Not really :-) It's more deep than that.
... You can dial in if you wish.
... We're running a bit behind schedule due to the very civilized French conception of what "dejuner" means :-) It ain't sandwiches in the meeting room!
... So the "WSA review" session probably won't start until 16:30 ... there's still the OWL discussion on the joint agenda

<mchampion> A "resource" is whatever fufills the contract specifies by the interfaces; it doesn't have to be physically identical or even a set of communicating interfaces, unless that is implied by the contract.
... A wsdl:service has a QName ... a "Resource" doesn't necessarily have a URI ... which is a "web service"?

<mchampion> http://ietf.org/rfc/rfc2396.txt
... ACTION: WSA will write up our own take on the Plato-Aristotle debate on the true meaning of "resource" -- concrete thing or shadow of it

<dbooth> I am interested in helping on this action item, having spent a fair amount of time on this question already.

<ugo> Hi Hugo, I am not sure whether my phone connection is not on,
... or the bridge is turned off

<ugo> how do I enter the passcode? is it *2123*? I tried 2123# but didn't work

<mitrepauld> I just tried to dial in, I entered the passcode and #, but I think I am on hold (music playing). Sound right?

<mchampion> For those who just joined, we are about 1.5 hours behind schedule

<mchampion> Bijan Parsia's presentation is on the web, URL to be posted eventually ...

<ugo> What is the plan for the rest of the day? We had 1.5 hours reserved for remote participants, but 45 minutes are already gone
... We didn't even get the OWL slides, so it's pretty difficult to follow anything

<ugo> Question on composition: any connection with existing choreography work at W3C and OASIS?
... Could somebody ask Bijan for me? (unless he already addressed that and I didn't hear)

<hugo> Martin: not talking for the WSCWG because they are still working on the mission statement:
... - BPEL4WS seems like representing one party's POV, hence the name orchestration
... - choreography represents the shared global view
... Joint session over

<hugo> yes
... you may want to propose to host it if this is a problem for you

<mitrepauld> Chris: focus on specs that comply with architecture, not an implementation that conforms.

<mitrepauld> Analogy to OSI 7 layers - TCP is sort of layer 4, IP layer 3, but may not strictly conform to all OSI layer 4/3 functions. OSI useful in providing a frame of reference for comparison, but not conformance.
... WSA should be used to relate specs to the boxes in the WSA stack diagram (Fig 3). SOAP fits in the Messages box, WSDL fits in the Description box, OASIS spec XYZ fits in box ???
... Gap analysis - boxes without specs - need WG to define spec.
... Specs associated with the Messages box shall ... define messages in terms of XML Infosets? SOAP 1.2 conforms, SOAP 1.1 does not.
... Terminology used in OASIS Spec XYZ conforms (does not conform) to same term as defined in WSA Concepts (or glosary?)

Web Services Architecture Face to Face Meeting
14 May 2003

Attendees

Contents

Summary of Action Items

Web Services Architecture Face to Face Meeting 14 May 2003

Attendees

Contents

Summary of Action Items

Web Services Architecture Face to Face Meeting
14 May 2003