The "change itinerary" sequence diagram has only 17 objects (lifelines, including the Client actor), but it is already very big in my opinion. But it spans 5x2 A4 portrait pages (using 8-10pt Arial fonts) already. I've included the JSPs as object, the front controller, as well as the transfer objects involved.
The "prepare itinerary" sequence diagram is even worse.
Since the example in Mark Cade's book also include the JSPs and their controller, I opt to do the same in the assignment. But then this raise the question, should I also draw the same diagram for the GUI clients? The interaction from the point of the UI's controller backwards will be mostly the same, being different only in the request/response paradigm of HTTP to the more interactive GUI-based systems.
My question is this now:
- Should I draw two sequence diagrams for every use case, for each type of client? Are GUI vs
JSP clients that different?
- Would it be appropriate to leave out the transfer objects in the sequence diagrams?
- Should I leave out the Home interface of EJBs in the process of creating a bean instance in the sequence diagrams?