- From: Ricky Ho <riho@cisco.com>
- Date: Mon, 03 Feb 2003 11:21:41 -0800
- To: Mayilraj Krishnan <mkrishna@cisco.com>, arkin@intalio.com
- Cc: public-ws-chor@w3.org
Mayilraj, thanks for your response. My comments are embedded ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- The external behavior can be modelled in both activity and transition diagram always. (1) State Transition Diagram Each action is taken when events are detected by the state machine. Action will get the inputs by the event. Here emphasis given more to the states of objects and transitions of the states. In practice UML recommendation is also to model the reactive objects(normally single) by state transition. <Ricky> From the perspective of "external behavior", I think the emphasis should be in "events", "transitions", "states". I'm not sure about "action" because it seems to be exposing private implementation of how a party respond to an event. </Ricky> (2) Activity Diagram Each action is taken when previous actions are completed or all the required inputs are available for the object. Here emphasis given the order in which the actions are taken. <Ricky> Following my above argument that "action" is private implementation details, then the emphasis in "the order of actions taken" is certainly undesirable from a "public behavior" perspective. </Ricky> That is the reason I view better to model the external (abstract) behavior using activity diagram. Notations: Activity Diagram State Transition ----------------------- ---------------------- Action Simple State Activity State fork/join fork/join transition transition <Ricky> I have a different opinion. It seems to me that an "ARC" in the state transition diagram should correspond to a "NODE" in the activity diagram. On the other hand, an "NODE" in the state transition diagram should correspond to an "ARC" in the activity diagram. </Ricky> You can not define the swimlanes (does not have semantics in UML) in state transition but the roles representation is clear using swimlanes and can be easily visualized in activity diagrams. <Ricky> I think it is easier to extend the "event definition" of the "state diagram". E.g. (Event is always "role X is sending a message M to role Y"). </Ricky> I guess I'm talking about the opposite. I think a "private process" is richer than a "public process". By taking away the "process variables", "conditions", a "private process" will become a "public process". Is that true ? Here I agree with Asraf comment "The fine line between public and private process is computation intensive nature". I think this computation intensive only you are calling richness. In that case yes, private process is richer than public process. But if it is not computation intensive I don't know why we have to model the operation using activity or state transition diagrams. If you remove the states, actions, transitions from the state machines it becomes flowchart. <Ricky> I don't understand here. What is left after removing states, actions, transitions from the state machines ? How come it turn into a flow chart ? An example ?? </Ricky> Best regards, Ricky
Received on Monday, 3 February 2003 14:22:33 UTC