Re: [soa-rm] What is SOA (Really???)

[Rex Brooks <rexb@starbourne.com>]

The most simple level, the atomic level, for me is one service and 
one service consumer. That also defines a community and it allows a 
full description of both service and service consumer, which together 
form the architecture. It is possible to model a service in 
isolation, but I would say that that case is not the one that 
interests us. It is not why we are here doing this. We are here to 
take the existing community and abstract it's most basic fundamental 
components in order to refine a reference model for SOA, As for who 
does the orienting, it is the community, which is well beyond clearly 
needing a model to guide future development of SOA which already 
exists in overabundance.

By building the set of basic components that will allow the more 
complete set of features Ken described yesterday in response to what 
makes SOA different from Distributed AD, we are really just taking 
what already exists and abstracting from that, regardless of the fact 
that we are attempting a top-down modeling effort. To some extent, we 
are also missing the argument about whether or not a model-driven 
architecture is the best direction for organizing this effort going 
forward. I think it is, but our socratic methodology insists we 
answer the question of why is a MDA better than, say Agile 
Methodology or Extreme Programming where everything is a special case 
and ought to be built around the specific existing situation.

Ciao,
Rex

At 4:30 PM +0200 5/27/05, Gregory A. Kohring wrote:
>You have not quite captured the debate. It is not that I feel features
>needed to make multiple services function are superfluous, it is just
>that no one has ever clearly said what those features are.
>
>At the abstract level, what concepts do you think are required?
>
>
>Take your community example and suppose its communication model is
>the Internet. Would apply such a model to a village of 3 houses,
>where people could just walk across the grass to talk with each
>other? The concept of a community is obviously more fundamental
>than such a communication model would allow, so do you need to mention
>it all? While the communication model is important, in my opinion it
>does not enter until you are ready to create a reference architecture
>for a particular type of community.
>
>As I see it, that is the problem we face. How to make a reference
>model simple enough that it applies to simple situations.
>
>
>-- Greg
>
>
>
>Don Flinn wrote:
>>  IMO the TC is spit into two camps and many times the two contingents are
>>  speaking past each other, enumerating their own view.  Rebekah's
>>  variation of the house analogy captured the difference:
>>
>>  A- One side looks at a Service Oriented Architecture from the viewpoint
>>  of the community whereas the architecture describes the houses
>>  (services) and their relationship to each other (coordination,
>>  choreography, etc.) and constructs their model from that viewpoint.
>>  B- The other side looks at a Service Oriented Architecture from the
>>  viewpoint of a single house (service) and constructs their model from
>>  that viewpoint.
>>
>>  Until and unless each viewpoint addresses the concerns of the other
>>  viewpoint we will never reach consensus. One can not understand another
>>  until you walk a mile in their shoes.
>>
>>  Following my suggestion, being of the (A) viewpoint, let me attempt an
>>  explanation of the (B) viewpoint.  B's contention is that the essence of
>>  what should be modeled is a service, where a service subsumes the
>>  service itself, Metadata and Discovery, Presence and Availability
>>  (Figure 1).  Once we have fully modeled a service, our customer, the
>>  specification writer, can develop a specification for any SOA
>>  architecture, including the complex scenario in Appendix B, by using the
>>  concepts of a single service multiple times, as needed.  Thus, features,
>>  which are exogenous to the service, that are needed to make multiple
>  > services function as a unit are superfluous to the model.
>>
>>  Does this capture the (B) view of what our RM should be?
>>
>>  Could a (B) viewpointer summarize the (A) viewpoints?
>>
>>  Don
>>
>>
>>
>>  On Fri, 2005-05-27 at 11:41 +0200, Gregory A. Kohring wrote:
>>
>>><quote>
>>>Make an example of something that is not conformant to the SOA RM and
>>>explain why.
>>></quote>
>>>
>>>
>>>One of the problems we are having in this respect is
>>>generalizing from the wrong basis model. Or more to the point,
>>>have we reached agreement upon what basis model SOA is generalizing
>>>from?
>>>
>>>In my opinion, SOA RM generalizes Client-Server; whereby
>>>the "client" is generalized to "consumer" and the "server" is
>>>generalized to "service". (In this sense, SOA is a fundamental model
>>>and we should try to keep it simple.)
>>>
>>>Seen from this viewpoint, we should ask what is the difference
>>>between client and consumer, server and service and the relationship
>>>between the respective pairs.
>>>
>>>A "client" has the server's description hard-wired. The policy,
>>>contract, data model and processing model are all hard coded into both
>>>the client and the server.
>>>
>>>A "consumer" on the other hand has some goal to achieve and must
>>>first discover a service which can achieve this goal, understand
>>>the service's policy and contract to see if the service's policy is
>>>in alignment with its own policy and constraints, examine the
>>>processing model to determine whether a session needs to be
>>>established before the request can be submitted and examine the
>>>data model to determine what format is needed for the input data;
>>>only then can the consumer submit a request to the service.
>>>
>>>If you accept this scenario (which I know is a big "IF" ;-), then
>>>an example of something which is Client-Server, but not SOA is
>>>FTP.  With FTP the policy (username-password authentication),
>>>contract (list of allowed commands), data model (byte order of the
>>>ftp packet) and processing model (control channel, data channel)
>>>are all hard-coded in both the client and the server, there is no room
>>>for dynamic inspection and negotiation.
>>>
>>>In my opinion, it is this inflexibility which forms the main
>>>demarcation between the Client-Server model and the SOA model.
>>>
>>>
>>>-- Greg
>>>
>>>
>>>
>>>
>
>
>--
>======================================================================
>G.A. Kohring
>C&C Research Laboratories, NEC Europe Ltd.
>======================================================================


-- 
Rex Brooks
President, CEO
Starbourne Communications Design
GeoAddress: 1361-A Addison
Berkeley, CA 94702
Tel: 510-849-2309