What if the composite exposes two services, one conversational and one non-conversational? Would the conversational service act as in 3 and the non-conversational service as in 4? That's not clear from your text. While it seems desireable to be able to answer yes to the second question, I'll observe that it means that components in that composite will run differently (different context) depending on the inbound service. This is another divergence from request scope.

...and while I have your attention...regrets for the call tomorrow.

Dave Booz STSM, SCA and WebSphere Architecture Co-Chair OASIS SCA-Policy TC "Distributed objects first, then world hunger" Poughkeepsie, NY (845)-435-6093 or 8-295-6093 e-mail:bo...@us.ibm.com http://washome.austin.ibm.com/xwiki/bin/view/SCA2Team/WebHome

"Michael Rowley" <mrow...@bea.com> To 03/12/2008 10:14 "Barack, Ron" <ron....@sap.com>, PM "OASIS Java" <sca...@lists.oasis-open.org> cc

Subject RE: [sca-j] ISSUE 21 - Clarify Request Scope lifetime

During the F2F it was observed that the request scope is similar to the conversation scope, except that it automatically propagates around the local components of a composite. Perhaps it would be simpler to take advantage of this similarity and also expand on conversation scopes. If an entire conversation can be marked as conversational, it could mean that all of the components within the composite work within the same conversation. This is also something that BEA has had customers ask for on its own right. However, if we did this, the request-scope could be removed, since basically same semantics could be achieved by having a composite that has been marked as “local” be also marked as conversational.

The rules that we discussed at the F2F were:

1. Composite can be marked to propagate conversations through all components within the composite ("propagatesConversation"). 2. Request scope goes away. 3. If the composite service has a conversational interface, then if the composite is marked as "propagatesConversation" then the conversation of the composite's client will be propagated throughout all of the components within the composite. 4. If the composite service has a non-conversational interface, then if the composite is marked as "propagatesConversation", then (unless step 3 applies) a new conversation will be started on each operation of the composite service and propagated throughout the components within the composite. 5. Marking a composite as "propagatesConversation" acts as if all the components have been marked as "propagatesConversation". Marking a component with "propagatesConversation" means that any conversationID passed into the component through a service will be passed with any call on a intra-composite wire from the component.

If this were to happen, it would probably have to be done in either the policy or assembly TC.

Michael

From: Barack, Ron [mailto:ron....@sap.com] Sent: Thursday, January 31, 2008 11:26 AM To: OASIS Java Subject: [sca-j] ISSUE 21 - Clarify Request Scope lifetime

http://www.osoa.org/jira/browse/JAVA-21

Von:: Michael Rowley [mailto:mrow...@bea.com] Gesendet: Montag, 21. Januar 2008 21:24 An: OASIS Java. Betreff: [sca-j] NEW ISSUE: Clarify Request Scope lifetime

RAISER: Michael Rowley

TARGET: SCA Java Component Implementation Specification section titled “Request Scope”

DESCRIPTION:

The section currently starts with the following sentence:

“The lifecycle of request scope extends from the point a request on a remotable interface enters the SCA runtime and a thread processes that request until the thread completes synchronously processing the request.”

From this description, it is not clear whether the request scope lasts through a remotable call to another component that happens to be local. In one possible interpretation it would depend on the binding. A call through a web service binding would be seen as changing threads, and therefore would be a new request scope. The same call through an SCA binding might be assumed to remain within the thread and therefore be within the same scope.

It is probably a bad idea for the scope to depend on the binding that is used, and it may even be a bad idea for the scope to depend on whether a call through a remotable interface _happens_ to be local.

PROPOSALS:

1) Have the request scope be only for a single remotable operation call. From that operation, any request scope component that is reached through only local-service calls would reach the same component instance. Calls through a remotable interface would introduce a new request scope.

2) Alternately, the request scope could last from the time a request “enters the SCA runtime” until it is done, but with the clarification that the “SCA Runtime” is domain-wide. As long as a call is made to another SCA component within the same domain (irrespective of the binding used) it is part of the same request scope.