WS Description WG Teleconference
9 Jan 2003

See also: IRC log

Attendees

Contents

• Agenda Items
  1. Approval of Minutes
  2. Review of Action Items
  3. Administrivia
  4. Publication issues
  5. Attachments
  6. Proposal: MEP support in operations
• Summary of Action Items

Approval of Minutes

Scribe: Minutes approved: http://lists.w3.org/Archives/Public/www-ws-desc/2003Jan/0002.html

Review of Action Items

Scribe: ACTION: 2002-11-12: Paco will write two options for naming faults: schema vs WSDL. -- DONE
... ACTION: 2002-11-12: Roberto will try and come up with another proposal for eliminating message, the discussion goes to email or the next f2f. -- PENDING
... ACTION: 2002-11-12: Glen and Paco to chase the Global Grid Forum WRT services implementing a single portType. -- DONE
ACTION: 2002-11-21: Don Mullen to detail changes/addition necessary to unify SOAP and WSDL MEPs. -- PENDING
... ACTION: 2002-11-21: Jonathan to refer R120 text to TAG, referencing TAG issue fragment inXML-28, when that text appears in the draft. -- PENDING
... ACTION: 2002-12-05 Glen to write up a description of the issues surrounding property description in WSDL. -- PENDING
... ACTION: 2002-12-05: JeffS to see what would be needed to publish his TCP binding on output only as a Note. -- RETIRED
... ACTION: 2002-12-19: Jonathan to find material proposing restricting services to a single portType. -- PENDING
... ACTION: 2002-12-19: JM to find volunteers for writing MEP's for input/output and output/input. -- PENDING
... ACTION: 2002-12-19: Jacek to write up text on SOAP response MEP after Don send his proposal for requrest/response MEP. -- PENDING

Administrivia

Scribe: ACTION: Glen to email Jonathan to schedule F2F time to discuss property description issues.

JMarsh: Please try to register for the F2F using your best guess. The meeting will end at noon, not at 5PM as stated on the registration page. Are there any additional topics for the F2F agenda? Steve Graham's operation naming is a good candidate (if we're not done today). Attachments too. Any other topics?

Scribe: No reply.

JMarsh: We could have a joint session with the ws-arch group. Suggested topic: what is the boundary of descriptions (as opposed to orchestration)?

JMarsh: Registration for the 3/2003 F2F is open. If there are any groups (e.g. xml core, xml schema) we want to meet with at the F2F, please send suggestions.

JMarsh: Also, no telcon on 1/23.

Publication issues

JMarsh:  what's the status of the draft?

Gudge: We could publish by the end of the month.

JMarsh: would like to publish a draft before the f2f.

Sanjiva: Shouldn't we incorporate decisions from the f2f if we publish after that?

Gudge: I can try to get a draft out by the end of next week.

Sanjiva: Got feedback that the spec is hard to read for non-infoset-aware people.

SteveG: was there any plan for a primer, like the XML Schema primer?

 dbooth: SteveG, we are doing a primer

JMarsh: Would like to keep spec up-to-date with the discussions.

Sanjiva: It's going to be the first draft that uses infoset.

SteveG: What is schedule for primer?

DBooth: The current primer draft is at http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/wsdl12/wsdl12-primer.html.

Gudge: Roberto, do you have any editing time next week?

JMarsh: Let's try to publish a draft as soon as we can.

JeffreyS: Part 1 is near-current, but part 2 isn't. would we publish part 2 as well?

JMarsh: What's on the editorial list for part 2?

Jean-Jacques: Features and MEPs will impact part 2. Todo list for part 2 has 3 items.

JMarsh: Shouldn't we then do those and publish parts 1 and 2 at the same time?

Jean-Jacques: OK.

Gudge/Sanjiva: OK to have something read by the end of next week.

JMarsh: SteveG can tells us why a new draft is needed.

SteveG: OGSI WG plans to release version 1.0 in February.
... Would like to use a WSDL 1.2 draft, but some people suggested using WSDL 1.1 plus additions.
... Drop-dead date is end of January.
... No option is ideal.
... With WSDL 1.1, we'd use GGF extensions and that creates legacy.
... Cannot wait indefinitely either.
... Commitment is to WSDL 1.2, with a timestamped draft being using in the interim while waiting for the recommendation.
... Using a timestamped version limits risk of deviation from the standard.

JMarsh: We already use timestamped URIs in our documents.

SteveG: open content model and portType inheritance are critical to us. Would like a draft before the f2f if possible.

JMarsh: GGF will have to decide whether the draft will suit their needs or not.

SteveG: Naming of operations is concerning us.

Scribe: ACTION: Editors to have a draft ready a week from tomorrow.

Attachments

Scribe: Requirement to support attachments: http://lists.w3.org/Archives/Public/www-ws-desc/2002Dec/0080.html

Jean-Jacques: XMLP WG working on abstract attachment feature.
... Thinking on producing a concrete attachment feature now.
... Discussing what to use (SOAP with attachments, DIME, something else).
... f2f may be too early.

Gudge: XMLP WG will discuss attachments at the plenary f2f.

Proposal: MEP support in operations

Scribe: http://lists.w3.org/Archives/Public/www-ws-desc/2002Nov/0085.html

Scribe: Sanjiva's <interaction> concept: http://lists.w3.org/Archives/Public/www-ws-desc/2002Dec/0073.html

JMarsh: Sanjiva proposed an <interaction> element.

Sanjiva: Use <interaction> rather than <operation> for general patterns.
...  <operation> would be syntactic sugar for one kind of interaction.

JMarsh: what's the advantage?

Sanjiva: General patterns may require different semantics for the ordering of parts, etc, than ordinary operations use.
... Or we could keep <operation> and say that the URI defines the style of the operation.

Gudge: Would <operation>s and <interaction>s map to the same thing in the component model?

Sanjiva: <operation> would be syntax for a certain kind of interaction.

DBooth: In an <operation>, the direction of a message is given by input/output. would the URI specify that?

Sanjiva: we could either say "direction is given one way, ordering, etc. by the URI" or "everything is described by the URI".

Glen: For MEPs, the meaning of URIs would not be directly machine readable.
... Rules for a MEP are up to you to understand.

Sanjiva: Proposal is to describe the general case and have a different syntax for the request-response special case.

Glen: Not sure we need to add new syntax for it.

Gudge: "interaction" would be the only concept, but there is a simplified syntax for the common case.

Glen: From the component model POV, even in WSDL 1.1., there is a concept of an interaction, but it wasn't crisply defined in the model as we are trying to do now.
... So we should recognize that such a thing exists.

Jean-Jacques: I support that.

JMarsh: So we should see if SOAP MEPs are adequate.

Gudge: What kind of axes can you describe or the interaction as a whole?  Synchronous/async?  In vs out?

Glen: Try to keep it simple.
... The spec for a MEP would describe everything, all you need to do in WSDL is tag the messages with the appropriate URIs given by the spec.

JeffreyS: +1 to DBooth.

DBooth: Describing message direction would be useful.
... It would help code generation.

Glen: it could be used by processors that don't know every MEP being used.

JeffreyS: if you don't understand a MEP, what do you really understand about the interaction?
... It seems you'd understand the data at least.

Glen: A MEP could be a "vote" to me, but a request/response to the nodes I ask to vote.

JeffreyS: Two WSDLs could be compatible even if they don't use the same MEPs.

DBooth: it's better if the information is explicit rather than hidden behind URIs.

JMarsh: Is having an explicit direction enough?

Sanjiva: I don't think it's a significant benefit.

DBooth: I think it is.

JeffreyS: It has significant benefits.
... For instance, intermediaries would benefit, think e.g. of a firewall.

DBooth: Monitoring applications would find it useful.

Sanjiva: I'm dropping my objection so we can get past this minor point ;-)

DBooth: Why shouldn't the information be explicit? one reason is if the description is too complicated.

JMarsh: Do we need anything besides an opaque URI for a MEP? (keeping the direction bit separate)

JeffreyS: Earlier proposal to have MEPs cover more than one node (and WSDL too).

Glen: A WSDL describes the point of view of one particular node.

Glen: Use case for MEPs: enable request-response over a transport that doesn't support it directly.
... There is a continuum from this to full orchestration.
... We could stop at some point in WSDL and then use the same concepts for orchestration.

JeffreyS: PortType A talks to two other portTypes, portType B does the same. what happens when I inherit from both?

Glen: This would be on the other side of the line for WSDL.
... You have URIs for each role that is involved, so in the example A would talk to roles X,Y and B to X,Z.

DBooth: Intermediaries fall in this category.

Glen: Intermediaries look like one node to both sides.

JMarsh: So direction is the only characteristic we express directly?

DBooth: Sequencing, at least when it's simple.

Sanjiva: Sequencing is definitely choreography.

JeffreyS: In Sanjiva's proposal, the well-known request-response from WSDL 1.1 would be covered.
... In the general case, we may not have sequencing (it's left to the MEP URI).

Glen: Don't like two constructs. couldn't we use one construct with defaults instead?

Sanjiva: In my proposal it's not a default: <operation> is defined in terms of interaction.

JMarsh: We're arguing over syntax now.

Gudge: Proposal that we work it out at the component level first.

JMarsh: So we design <interaction> first, work on unifying it with <operation> later.

Gudge: Overall, we agree on the component model, disagree on how to come up with the syntax.

JMarsh: Can we compare syntaxes at the f2f?

Gudge: I can work with Sanjiva on the component model.

JMarsh: we seem to agree on the abstract part, so we can write it down in the spec.

Glen: I'd like to see both proposals (new syntax vs. current one).

JMarsh: can the syntaxes for <operation> and <interaction> be unified or not?

Scribe: ACTION: Gudge and Sanjiva to prepare in time for the f2f a draft of the abstract model for interactions.

Scribe: ACTION: Sanjiva to write an email to describe the two proposals for the syntax and their mapping to the abstract model.

JMarsh: Outstanding question: how do these MEPs relate to SOAP MEPs?

Summary of Action Items