URN meeting, July 22, 2008
University of Toronto, Canada
Participants: Daniel Amyot, Jordi Cabot, Golnaz Elahi, Jennifer Horkoff, Jose-Norberto Mazón, Eric Yu
- In Z.151, constraints specifying that a metaclass is essentially abstract should be spelled out as “All intances of must appear in one of its subclasses (that is is abstract).
- In Z.111 (and indirectly in Z.151), we must specify that multiple classifications are not allowed. For instance, a GRL link must not be allowed to be a decomposition and a contribution at the same time.
- URN link should likely be a URNmodelElement.
- LinkRef should have an attribute to distinguish straight lines from curve lines.
- The showAsMeansEnd attribute of ConcreteURNspec should be moved to GRLspec
- The consistency of qualitative/quantative values will be left to tools to check, if needed.
- IOR and XOR decomposition will be kept.
- Strategies look OK. Will include 2 examples of analysis (informative) in the standard.
- Intentional element references should have a size
Issues that need further/special attention:
- There will be different ways to interpret the "importance" of intentional elements.
- Several restrictions on the current definition of dependencies should be relaxed. In particular, the name of an intentional element should be unique inside an actor (and unique outside any actor), and not necessarily inside the entire model. This would allow for diagrams like Goal(inside Actor1) --D-- Goal --D-- Goal(inside Actor2), where the three “Goal” are different definitions.
- The term "evaluation strategy" sounds much like an algorithmic procedure (and not like initializations) and hence should be explained carefully.
- Actors should perhaps be allowed to be included in many other actors. This would lead to visualization issues however.
- Actor containment, as currently visualized, might give the impression that the intentions of the sub-actor are shared by the parent actor, which is not the case.
- Having a circle for actors on top of the actor boundary (as in OpenOME) will help dealing with SD models and actor collapsing.
- The current metamodel does not support well the situation where the modeller builds the SD model first and then expands the actors to allocate the dependencies to intentional elements inside the actors.
- Should dependencies be allowed between an actor and its sub-actors (and/or vice versa)?
- 25 Jul 2008