[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: draft f2f minutes
_________________________________________________________________ Martin Chapman Consulting Member of Technical Staff Oracle P: +353 87 687 6654 e: martin.chapman@oracle.com
23 members, quorum is 12 Roll call: Eric Newcomer Doug Bounting Jan Alexander Martin Chapman Mark Little Bob Haugen Greg Pavlik Bryan Murray Jeff Mischkinsky Joseph Chiusano Malik Saheb (phone) Dale Moberg Peter Furniss (phone) Meeting scribes were Jan Alexander, Mark Little and Greg Pavlik Agenda review: Basically going through the issue list Bob: Can one WS-Context interoperate with multiple ALS - how that works? Bob: Can existing WS-Context implementation work with new ALS it didn't about previously? Martin: we are not officially quorum, so let's proceed and write the issues down and vote about them when other will join us to have a quorum. Martin: Have to join BPEL call, propose to have a break and to continue at 10:30 with the current editor's versions of the documents, hopefully we will have quorum at that time. Eric: Should we go through the complete editor's versions, or just walk through the existing editorial issues and discuss those only? Eric: Documents were updated about 61 issues, either editorial or voted on during the TC concalls. Propose that these would be approved. Motion: To adopt editor draft 01 dated on 21.April as a working draft for the purpose for discussing other issues. The motion is seconded by Mark Little. Eric: The dafts of the model diagrams were posted at least twice to the mailing list, so we assume they are discussed and understood, so we would like to make this an editorial issue. [Add a URL to the Eric's mail summarizing the issues, that were fixed (mail subject: Issues fixed in editor's spec) from the April 18.)] Mark: Only difference between 01 and 02 drafts is restructuring of the documents and incorporating the new model from the mailing list. Martin: Let's discuss the open issues now. Issue 40: Remove restriction from the schema and establish convention for the hierarchical names of the status. Issue 46: This should be left to the implementation, because there is multiple ways how to do that and we don't want to restrict to the implementation of this. Mark: Motion: Add an implementation note in the text to setContent: Concurrency control of a context passed by a reference is an implementation issue. Motion seconded by Eric. Issue 50: actually two issues. The missing piece can be either added to the WSDL or removed from the spec. Mark: Motion: Remove addIdentity and removeIdentity from the spec of the ALS portType. Seconded by Greg Issue 59: let's wait for the rest of issues to be resolved (we need model resolved for it). Leave open, it is editorial. Issue 60: Mark: Motion: Activity is an abstract thing, context allows you to capture activity, the maintainer of the context should be called Context Service, not Activity Service. Seconded by Greg. Martin: we should change the name Context for the marketing reasons to Cookie. There is large overlap between Context and Cookie in HTTP. Eric: Let's postpone this after the issues discussion. Jan: Shouldn't we rename ALS to CLS (Context Lifecycle Service) if the activity sevice is renamed to context service. Erik: Yes, ALS should be CLS Greg: Is it possible to introduce CLS/ALS into the context by application of the context - coordination framework? Is the CLS/ALS in the WS-Context essential? If it is optional for the context and required for the coordination, shouldn't it be introduced by WS-CF? Mark: We need it for the interoperable context augmentation. Martin: Let's make it a new issue. Issue 61: Greg: originally there were two ways how ALS/CLS said I'm associated with specific protocol, but that is another issue 106. So close this issue, because it was obsolete by issue 106. Issue 91: Leave it open for now, Greg will provide a text. Issue 94: Mark: Suggestion is to keep setContext simple now and when enough implementation experience will be gathered, we can return to this. Motion: close it as a non-issue. Seconded by Greg. Issue 109: Mark: clarification text will be added, editorial issue. Issue 123: Martin: Dough sent an email proposing that the implementation should decide, if to pass context by reference or by value, we should make it out of the scope of the WS-Context specification. Mark: we leave the signature of the begin() method as it is and maybe add some text explaining the possible return values of the method. Jan: should we add something like: "Implementation of the context service decides, whether it will return reference or value of the context as the response of the begin() method". Martin: Add to it, that it is always possible to convert value to the reference and vice versa. Bob: Returning the questions from the beginning: Can a context implementation interoperate with ad-hoc ALS/CLS or have it know from the beginning the referencing specifications of those CLS/ALS? Mark: I think this was resolved in the new restructured editorial documents. Lunch break. Restart after lunch looked at Bob's issues: (i) can one WS-CTX implementation interoperate with ad hoc ALS implementations or does it require knowledge of the referencing specs? (ii) can multiple ALSs be involved in the same context, or can they be associated with the same context service? Mark's answer is that a context implementation should be able to use any ALS as long as it conforms to the WSDL. Also one ALS per activity is now required. Greg: where does it state that one activity per ALS? Mark: enlistALS Greg: there is an inconsistency between the start of the spec. and that section. Martin: editorial issue to tidy up. Bob: one context to one activity. But there may be different snapshots of that context. An activity can have one or more contexts representing it, but each instance shares the same context identifier. Bob: one activity not one context? Is the intent that the main identity is the activity. Greg: yes, the context is just the wire-level representation of the activity. Eric: the activity is the "execution environment". Martin: we'll revise the UML diagram and put it into the text. Must make sure that the text syncs up with the diagram. This only occurs when passing by value. Model should capture the fact that multiple protocol implementations can be wrapped by the single ALS. Context is a view of an activity. Can have multiple views during time. Issue 11: not resolved. Simeon to send correct WSDL. May have to re-assign (Greg to check). Issue 47: not sure about the issue. Choices could be boiler plate saying our specs are composable with WS-Security, enumerate potential threats for using WS-Context without appropriate security. Is security of the context service implicitly securing the activity? Likewise, if we don't secure the context service and a security ALS is added, does that secure the context service? Action point: do we need to change out schema in context inorder to sign these within headers? Martin to check. We need a dedicated teleconference to discuss this. Issue 48: text clarification required. Editorial. Issue 52: define the completion type for context spec. but remove all values from the specification. State that the referencing specifications can define their own values for this type. Remove completeWithStatus and make complete take an optional completion status value (min=0, max=1). Each ALS is allowed to interpret the receipt of a complete message with no parameter in whichever way it requires. The referencing specs. can define the default value. Mark makes motion. Dale seconds. Issue 53: close no change. All delegated to ALS. Mark makes motion. Jan seconds. Issue 54: close no change. Mark makes motion. Greg seconds. Issue 62: fixed; in editors copy. The change to incorporate the new model text removed the ALS-configuration id in favour of the protocol-type and is in the AssertionWithProtocolURIType. Mark makes motion. Dale seconds. Issue 72: editorial. Close issue. Should go into the Primer. Greg makes motion. Mark seconds. Issue 92: closed as resolved by issue 91. Greg makes motion. Jan seconds. Issue 93: close issue. No action. Mark makes motion. Greg seconds. Issue 99: leave open. Action point: Martin to determine what best practice is. Issue 100: closed because revised model uses a single ALS per activity. Greg makes motion. Mark seconds. Issue 101: closed because of issue 52. Jan makes motion. Greg seconds. Issue 114: closed because of issue 52. Mark makes motion. Greg seconds. Discussion topics: (i) addressing (WS-MD). (ii) state transitions (state model). (iii) ALS - can it be collapsed into context service. --------------------------------------------------------------------------------------------------------------------------- April 29 2004 (Greg Pavlik, scribe) Agenda for the day: general issues discussions OASIS namespace protocol versioning addressing/referencing state transitions ALS - can it be collapsed into context service doug to talk about model document planning implementation group Implementation subgroup presentation by Greg to post ppt show to TC site Jan expressed interest in a general interop event AI for Eric: find place to do publicity demonstration General Issues of which there are 3 Issue 49 assigned to Simeon Should we use substitution groups? Resolution: no. Fix produced, to be incorporated in next draft Closed, resolve later Issue 124 should we include synchronous-style port types? defer until message delivery discussion Issue 125 Respondant needs to be changed throughout all specs Editorial change for CF and TXM Resolution, make sure ResponseHandler is used throughout. Discussions OASIS namespace Doug: multiple answers, there is no centrally managed namespace. However, there is a long history of TCs using http://www.oasis-open.org/committees/short for of committee name/[documents|schemas] There is also an RFC for a URN scheme. People like resolvable namespace identifier. Eric, what about webservicestransactions.org? Doug: several issues with that. AI: chairs to find fixed location on OASIS site Protocol Versioning Mark concerned about versioning being addressed now, under assumption that changes may occur after WS-Context is adopted but before other specs expose problems Martin doesn't want to have to rev up through OASIS level multiple times; and no errata process. Martin: every new version is in a new namespace. General agreement; Dale suggesting version attribute for "tweaks" on final spec. State transitions Do we need state machines for WS-Context? Resolved, this is not significant in light of specification changes removing failure states. Doug Model Issues Suggestion that Context be treated as a shared information space. We also talk about the hierarchy of contexts, but not of a linked chain of various temporal contexts. Not comfortable with name protocol type -- could be misconstrued, and ought to be type. Doug motion to change protocol-type in Context structure to "type". This has ripple effect to all use cases of protocol-type. Motion carried. ALS discussion Mark: delegate much to ALS, can it be removed as an independent entity? Martin asks if there exists a market for separate ALSes? Mark: possible to add it back at later date, but removal simplifies and removes adoption hurdle. Doug: question of genericity? do more use cases exist beyond coordination? Motion from Greg: Motion to remove ALS and validate the Context Service definition is complete. Mark seconds. Discussion ensues. Motion carried. Referencing/Addressing/Request-Response style Do we want to use w3c MessageDelivery for referencing services? Concern over status as a note. SOAP for example was adopted by Working Group Eric hesitant to tie spec to something in state Note. Discussion to provide open content model for references to web services and conformance claims around a subset. Motion by Jan to adopt an open content model for all uses of references to web services in the specification. Mark seconds. Motion carried. Discussion about addressing headers. Motion by Greg: Addressing scheme will be defined by association to reference model. If reference model does not provide addressing headers, an implementation will provide in a proprietary manner. Second Jan. Discussion, Eric expresses approval. Motion carried. As a result of this motion, the AssertionType addressing elements will be removed AI: editors will add to text to explain content model. AI: Martin will explore how we identify WS-Reference compliant implementations in the CAF content model. Request-Response WSDL Leave as is, need to initiate investigation on MD callback pattern. Issue remains open. Peter would like to see Logical request-response table. Liason slot Two new use cases for CAF that are not transactional ebMS (Messaging) Ian Jones chair of the ebMS group visiting to see if there is synergy between the requirements of ebMS group and WS-Context. Strong interest on both sides. ebMS looking to combine reliability, security and context from OASIS. Peter in now in attendance. Monica for BPSS/EBPP 2.0: multi-party collaboration, coordination and shared context becomes important. How to associate multiparty collaboration? Potential customer. In both cases, a draft around June to TCs for review. Scheduling Meetings Next F2F Oracle Redwood Shores July 13-15 AI for Jan: check on September, October F2F in Prague Next telecon starts May 10, usual time/call-in document schedule: from july we dont touch context unless we must (take it to cd) and we propose to oasis standard once we asses its stable.
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]