Here's an attempt of a simple UML-like diagram of these core concepts:
Dotted lines are optional. For instance, a Thing might be Derived from a Thing even if we don't know which Process Execution made the derivation.
Stuff is out of scope for the PIL assertion language, but forms the basis for assertions about Things. The Asserter is implicit in all statements and is therefore not linked in. Time has not been defined, but is optionally involved with almost all parts of the model. Version and Collections are not yet defined, but suggested for discussion for the F2F.
Image generated in OmniGraffle by Stian Soiland-Reyes. source
--Stian Soiland-Reyes 21:26, 5 July 2011 (UTC)
Thing now called 'Bob' until finding a better name. Stuff now called entity (F2F decission 2011-07-06)
- Is the process an assertion of BOB, a derivation of BOB, or either?
- What is an activity?
- generates implies creating. Can the process be a retrieval of BOB? If this is the case, would it be the generated BOB a new BOB or an existing BOB?
- I think we should note that a process execution is an entity state itself, i.e. it has invariant properties for which can be part of provenance (SMiles)
Note the issue raised at http://www.w3.org/2011/prov/track/issues/26 with questions about how to use use in a simple example.
comments from Deborah
- Agent is a class
- One can be stated to be an agent COMPLETELY independent of performing any action or participating in any activity. (supporting Paul’s statement).
- Independently there may be potentially multiple sufficient conditions for membership in the class agent.
- One such example of a sufficient condition for membership in the class agent: If something performs an particular kinds of activities (such as a process execution), then it will be inferred to be an Agent.
- In addition I would like agents to be allowed to be organizations, software programs, sensors