[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [soa-rm-ra] reaching closure on Action
Colleagues,
OK. I'm going to pretend to be colorblind as Mary
suggested and I'm going to apply one of our core architectural principles---the
principle of parsimony.
Here's what I see are the essential core concepts around
Action and their relationships (textual description only):
* Action (defined as before, cf, Lines 754-755 of PRD1)
but emphasize this is Separate Action.
* Joint Action (defined as before, cf, Lines 810-812 of PRD1) but move defn to follow Action. ARGUMENT (in favor of Joint Action): In SOA-based
systems, a Service doesn't "act" unless "acted upon". This is
why the concept of Joint Action is so
important.
* Drop the concept of Communicative Action (CA) in favor of
Interaction.
ARGUMENT: The latter (Interaction) is already a core
concept in the RM and, in my opinion, all we're doing by introducing yet
another abstract concept like CA is adding gratuitous
complexity.
* Interaction facilitates Joint Action
* Interaction (we've agreed for this RA) is accomplished via
Message Exchange
* Messages are the medium of Interaction between a Consumer
Agent and a Provider Agent (i.e., Service).
* Actions (and Events) are conveyed through
Messages.
CLARIFICATION: Actions are not carried out by Messages,
but the intent of Actions is embodied in the content of
Messages. Later, after we get closure on Action, we need to expand this
notion to include Events.
* Interaction is dependent on the visibility of a Service and
its interface (i.e., Service Visibility and Service
Interface).
* The Action Model (which is part of the Service Interface)
describes, in a manner that can be used by a Consumer Agent,
the permissible actions that can be invoked against the service. It serves a purpose analogous to a function or method
signature.
* Message Exchange Patterns (MEPs) are defined as part of the
Process Model (which, as with the Action Model, is part of the Service
Interface).
ARGUMENT: MEPs are temporal in nature. With the
exception of my proposed visual model for Service Interface, I have yet to see
any of the visual models proposed thus far capture this fact.
Alright friends. That should be enough for now.
Today, I'll volunteer to take notes since I'll be on travel overseas attending
an INCOSE event for ten days, beginning on Friday. I'll do my best to call
in remotely but no guarantee.
Happy architecting!
- Jeff E., JPL
|
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]