Issue 16552: Underspecification of initiating/non-initiating participants in non-trivial choreographies (bpmn2-rtf) Source: (, ) Nature: Clarification Severity: Significant Summary: Resolution: Revised Text: Actions taken: September 15, 2011: received issue Discussion: End of Annotations:===== m: webmaster@omg.org Date: 15 Sep 2011 08:19:12 -0400 To: Subject: Issue/Bug Report ******************************************************************************* Name: Michele Mancioppi Employer: IAAS, University of Stuttgart mailFrom: michele.mancioppi@iaas.uni-stuttgart.de Terms_Agreement: I agree Specification: Business Process Model and Notation (BPMN) Section: 11 FormalNumber: formal/2011-01-03 Version: 2.0 Doc_Year: 2011 Doc_Month: January Doc_Day: 03 Page: 315 Title: Underspecification of initiating/non-initiating participants in non-trivial choreographies Nature: Clarification Severity: Significant CODE: 3TMw8 B1: Report Issue Description: Page 328 reports the following text on non-initiating participants: "The Participant Band of the Participant that does not initiate the interaction MUST be shaded with a light fill." This wording does not cover some corner cases. Consider for example a sub-choreography with a start-event, connected to a parallel gateway which branches to two different choreography tasks, each of the latter initiated by a different participant. In such a sub-choreography it is unknown until enactment-time which participant is the initiator. But this is only the symptom of a wider-reaching problem affecting choreographies, sub-choreographies and (by reflection) call-choreography activities. When there is no choreography task that dominates of all the others, or worse, there is a race condition, modelers have no way to pick which participant is marked as the initiator of the (sub)choreographies. At first analysis, there are two possible, mutually exclusive solutions for solving this issue: 1) Additional constraints are specified for choreographies so that no such corner case can occur. However, this is very likely to result in not being able to model with BPMN 2.0 choreographies some inter-organizational processes that can instead be modeled with BPMN 2.0 Orchestrations. 2) Drop the differentiation between initiator and non-initiator participants in sub-choreographies. We see no real shortcoming resulting from this approach. In particular, with respect to the enactability of choreographies, knowing which participant is the first to act in a sub-choreography gives no guarantees as to the fact that the same participant will also be involved in the .last. choreography activities to be executed in that sub-choreography. Therefore, we can extract no useful information from it with respect to the enactability of what follows that sub-choreography. This is particularly true in the case of collapsed sub-choreographies.