OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

soa-rm-ra message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]

Subject: Re: [soa-rm-ra] Interacting w/Services Model Updated (a start on EDA & event-driven SOA)

Frank - some comments inline - Ken

On Mar 24, 2007, at 3:30 PM, Francis McCabe wrote:

 I think that the way to think of this is that a message exchange is one level of a joint action - it requires two (or more) participants. Each participant performs operations needed to support its participation (sending a message, receiving the message).

Then is there an action that is not a joint action?

 A long time ago, Tony Hoare talked about *synchronization* being the most primitive of constructs needed for distributed computing. This is similar in spirit to the joint action.
 Your dictionary definition of denote is on the mark here. In our message oriented architecture, we use messages to denote actions and to denote events.

I'm not comfortable with this.  The message doesn't *indicate* an action but rather it passes information that I may expect results in that action.  Now I could say a message denotes and event because I could say the message indicates the event has occurred.  But again, this emphasizes the difference between a message relation to an action and a message relation to an event.

 Herbert Clark talks about ladders of action; paraphrasing we would get:
Level 1: exchanging a well formed bag of bits;
Level 2: the message is about an action
Level 3: the action is one that is supposed to be performed by one of the partners
Level 4: some service is being requested by one participant of another
 He also talks about upward cause and downward evidence:
Level 4: the desired RWE is observed, which means that
Level 3: the action was appropriately requested, which means that
Level 2: the message was about an action, which means that
Level 1: the bag of bits was correctly formed.

Message exchange takes place at Level 1 here.

No problem here:

participant 1 --- level 4 --- level 3 --- level 2 --- level 1 ===wire=== level 1 --- level 2 --- level 3 --- level 4 --- participant 2

However, I'm not sure this level of detail is appropriate for the RA.

A similar train of thinking can explain how messages figure in events, and how events figure in observing RWE.


On Mar 24, 2007, at 9:02 AM, Ken Laskey wrote:

Catching up on email traffic and Jeff's additions to Interaction.

I missed the live joint action discussion, so I can use some clarification.  Is a request-response considered a joint action because I have two participants and by definition of the MEP I need both the request and the response to qualify as the resulting RWE?  Is everything a joint action except for me sending a message to which I expect no response?

Most of the material Jeff added about joint action applies equally to (simple?) action.  However, my processing of the words in the wiki gets the sense that in a joint action a (some?) service has more of an initiating role.  Am I reading in more than is intended?

Also, saying a message "denotes" both an action and an event seems like an overload for very different message intents.  The online Free Dictionary gives the following definitions for denote:

1. To mark; indicate: a frown that denoted increasing impatience.
2. To serve as a symbol or name for the meaning of; signify: A flashing yellow light denotes caution.
3. To signify directly; refer to specifically.

Does UML have a more precise meaning in context for denote?  Otherwise, the connection between message and action seems to be invoke (and in the reverse direction, Action is realized by Message); with respect to Events, it appears a Message notifies a participant of an Event, and an Event enters a participant's awareness through a notifying Message.  (I was going to say "becomes visible" but I didn't want to confuse with more inclusive Visibility under RM.)

We've tried to differentiate between Event and RWE but other discussions in this thread seem to say that Event reporting RWE is too limiting for Event.  To me, Event connotes something happening and is a subset of RWE.  So RWE could be the Event of sending a command to decrement an inventory and the final decremented inventory.

Minor question: is RWE a set of effects or each individual effect or both?  Is there a need to differentiate?


On Mar 19, 2007, at 1:40 PM, Francis McCabe wrote:

 You have misunderstood the connection that we looked at before here.

 An event cannot cause anything. Least of all an action.

 By definition, an event is a report f something that has happened that is of interest to someone.

 An action is the application of intent (I personally prefer the term force over intent, but that riases more potential for misunderstanding) on a target to achieve an effect.

 Actions may (but then again may not if the action was unsuccessful) give rise to events. An event may be interpreted by an agent as a reason for initiating actions.

 Messages may be used to denote events and actions. More accurately:

We use message exchange as the realization of communication of actions and events.

 More accurately still, following from Herbert Clark's notion of joint action:

Participants send and receive messages (which are inherently individual actions) in order to participate in joint actions (the exchange of a message, the performance of an action, the reporting of an event).

In our architecture, actions and events are both realized via message exchange. A message may be used to denote an action and/or an event.


On Mar 18, 2007, at 3:13 PM, michael.poulin@uk.fid-intl.com wrote:

Replying to the Jeff's question about relationship between Event and Action, let me suggest that an Event can cause an Action directly or indirectly; plus, in some cases, one Action might be not enough to cause particular Action. All this looks complex.

Would it be easier to define that an Event can relate to the Action via a Message? On the UML diagram, the 'denote' relationships might be bi-directional... In other words, I propose using following UML construct here:

      :EVENT  ------x-------->  :ACTION
- Michael

Ken Laskey
MITRE Corporation, M/S H305     phone:  703-983-7934
7515 Colshire Drive                        fax:        703-983-1379
McLean VA 22102-7508

Ken Laskey
MITRE Corporation, M/S H305     phone:  703-983-7934
7515 Colshire Drive                        fax:        703-983-1379
McLean VA 22102-7508

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]