The following topics are subjects for future work:
Validation of HIT models
The HITvis tool is able to visualize the runtime execution of valid HIT models only. For now, the validity of a HIT model has to be manually checked by the designer. Automatic validity check is a subject for future work.
A HIT model is valid if it satisfies the following conditions.
Each state has a name which is a unique state name in the model.
Two states are either disjoint or one state contains another one. Overlapping states are disallowed.
Each atomic state is contained in at least one non-atomic state.
Each non-atomic state has at least one start state and at least one end state.
Each transition has a non-empty label.
Each transition label has at most one literal (specifying the event which triggers the transition) and an arbitrary number of constraints.
If a literal in a transition label is negated, temporal constraints in the label must specify the time interval during which the non-occurrence of the event is required for the transition to fire.
A transition label may not contain a reference to an undefined variable or identifier in a process.
All concurrent processes are differentiated by process identifiers (i.e., the instantiating transitions of a concurrent process define the process identifiers).
Two transitions with the same source state, target state and label are disallowed.
A cycle of transitions which are not triggered by events is prohibited.
Queries within HIT states
Succinct specification of expressive stream processing queries will be achieved by formulating the queries within the states of HIT during which they are satisfiable. States provide valuable context for launching query execution.
Complex predicate-based conditions expressing the evaluation time of queries can be omitted in these queries if their evaluation time corresponds to state validity time.
A query is suspended as long as its respective state is not active.
The workow-determined conditions can be omitted in queries since their satisfaction in implicitly achieved by HIT.
The workow-determined conditions are shared among all queries and evaluated only once.
A query can access the summaries of processes.
Stream verification, dynamic process analysis, management and optimization will leverage the HIT technology thanks to the following built-in actions:
Default actions log the beginning and end of each application state of each process for later verification, fraud confirmation and process optimization.
Customized actions can override the default actions by deactivating, extending or changing the logging.
Extended actions further enhance the capabilities of workow monitoring by adding emitters that send events to processes which implement monitoring related decisions. For example, in an auction, if a bid event arrives in state I4 or I5, the bid is late and a rejection notification is sent to the bidder who placed it.
Sophisticated actions are based on statistics and analytics of process monitoring.
For example, in an auction, identification of popular items, high bidders and their purchase behavior can be exploited for making a profit. Examples of sophisticated actions are the dynamic determination of items and their prices depending on their recent popularity or discounted offers of current items of interest to high bidders.