Colleagues,
Attached is a copy of the tutorial Frank presented for us this
morning as part of a SOA tutorial series being internally
sponsored. It was excellent and thanks Frank!
Note there are some updated visual models, particularly for the
service ecosystem view. We’ll want to discuss these as we go
through Section 3 issues against the PRD 1 and, of course, discussion points to
be addressed during our forthcoming face-to-face.
Also, below is a message from one of our principal engineers who
is very active in OMG working the Executable UML Foundation spec and emerging
standard with Stephen Mellor et al. He’s the fella who was asking
all the questions Frank. In any event, I will ask him to provide some
feedback once we go to PRD 2.
Cheers!
- Jeff
From: Nicolas Rouquette
[mailto:nicolas.rouquette@jpl.nasa.gov]
Sent: Tuesday, October 28, 2008 10:52 AM
To: Jeffrey A. Estefan
Subject: SOA-RA comments.
Jeff,
Please pass along this feedback to the SOA-RA group.
1) accidental avalanche vs. intentional avalanche.
Could we use SOA-RA to specify an insurance policy providing coverage for the
real-world effects of the former type of accident but not the latter? If we
can, how would we use the SOA-RA policy specification to determine whether a
particular avalanche incident was factually accidental or incidental?
2) UML class diagrams help but are not enough to communicate the conceptual
distinctions and ontological commitments made in organizing the SOA-RA model.
For example:
- slide 8: Acting in a social context.
I suggest looking at JPL's State Analysis methodology for possible ideas for
explicitly characterizing:
- what a measurable state of the world is (i.e., a state variable
characterizing the state of some item in the real world)
- how a measurable state of the world relates to a goal (i.e., it is a
constraint on state)
- how actions affect the real world (i.e., actions coordinate the execution of
commands)
- how actors perceive the real-world effect of actions performed (i.e., using
measurements)
In this sense, an SOA is a society of distributed, loosely-coupled
goal-oriented control systems.
- slide 10: joint action vs. action
It seems that if you distinguish between action & joint action, then there
should be a distinction between intent and joint intention. Do those
distinctions really matter at this very high level of abstraction? I'm not
sure. I think that it might be useful to replace joint action with a more
general notion of goal-elaborated plan of action. If the model allows for
actors to have shared goals, then a joint action could be explained in the
context of a plan elaborating how the actors involved either coordinate or
compete in achieving the goal intent.
- using UML class diagrams.
Class diagrams are useful but it is very hard to get a grasp of the underlying
ontology of concepts and relationships.
It is also difficult to get a sense of the different kinds of behaviors that
are in scope in the SOA-RA.
Most of these behaviors are described in plain English in the subtext of the
slides.
At minimum, it is important to have a catalog of all of the behaviors that
pertain to the SOA-RA model.
-- Nicolas.