This is the workflow as it is today. The user starts out in the initial state on the left. Then there's an AND-split, a transition that triggers automatically, allowing the next two steps to proceed concurrently (or at least in any order).
The email verification transition is external (hence the letter, because the workflow system receives a message that it has occured), The admin approval is a user action, i.e. something a user must do. After both of these activities have taken place, the automatic AND-join transition can occur, and the user moves into authorized state.
The rest, we probably wouldn't use workflows for. We'd let the workflow end in authorized. Banning the user is something that happens outside the workflow and can happen at any time. But as modeled here, the administrator can ban the user, thus moving him into the banned state, from which he can be unbanned.
We'd probably also add an OR-split/join around the email verification transition, with a timed transition that occurs when the user hasn't verified his account after a certain amount of time, ending the workflow prematurely.
All workflows have a higher-level workflow associated with them, which allows us to cancel or suspend the entire case at any time. This would be used for denying a user access to the system. While there are extensions with hierarchy to Petri Nets that would be the conceptual model behind this, we won't do it this way. What we do is use a finite-state machine with states 'initiated', 'running', 'suspended', 'canceled', 'completed'. That will be sufficient for our purposes, while not burning any bridges for moving to the right thing later on.
The advantage of using a generic workflow subsystem for this is that it allows for easy customization. Siemens, for instance, have pretty tight security measures for approving users, sometimes requiring the user to go over his information again before allowing access. The above workflow could easily be extended with this functionality. Also, if, say, several independent departments need to be involved in approving users, the workflow subsystem would be what coordinates that effort.
This workflow starts with the user specifying his problem, thus putting the workflow into the open state. Once here, it will pop up on the responsive user's to-do list (sometimes called an in-basket), prompting for a fix. This user looks at the case and determines whether he can fix it or if he needs more info. This determines what new state, the workflow enters. You know the rest.
As before, the workflow system by default supports cancel and postpone actions, determining the overall state of the case. Since the workflow subsystem is what coordinates the effort, it would also be the workflow subsystem that takes handles communication between participants. In other words, the user responsible for performing an action on a case, can put in a message on the case at any time, with or without a workflow action being performed. All systems using the workflow subsystem would benefit from that.
| Last Modified: $Date: 2001/11/19 18:31:50 $ |