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

 


Help: OASIS Mailing Lists Help | MarkMail Help

soa-rm message

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


Subject: Re: [soa-rm] Types of Services (RE: [soa-rm] Definition of"Service Consumer")


Christopher,

My ultimate point here is that it is not SOA-RM's mission to decide what implementation paths are valid for SOA. I certainly would not develop a DLL based SOA, but that is just because doing something like that is unlikely to cross into my world as a architect of (mostly) Java software.

SOA-RM's mission is to provide a reference model that describes what, at a high level, an SOA is.

-Matt

On 11-Apr-05, at 9:41 AM, Christopher Bashioum wrote:

Matt,
 
this is an excellent argument.  This is getting at some really fundamental issues.  I don't know the answer, but I am learning a lot from this. 
 
One of the problems with the dll approach is that configuration management across an enterprise (actually, even on a particular machine) becomes a nightmare.  dlls can be (and often are) overwritten when a new software application is installed.   Because the application creator packages all the dlls that might be needed with their install - necessary to make sure their application will work.  However, other applications on the machine that use the same dlls that were just overwritten may not work properly because the new dlls don't behave exactly like the old ones (or, as is more often the case, the application package ends up overwriting a newer dll with an older version)
 
I thought this was contrary to what we were trying to accomplish via SOA.  All that being said, however, I can see your point that dlls have something in common with SOA.  Maybe we can agree that dll is "close to an SOA", but may be missing some fundamental items that would make it completely SOA.
 
 
From: Matthew MacKenzie [mailto:mattm@adobe.com]
Sent: Monday, April 11, 2005 12:16 PM
To: Christopher Bashioum
Cc: soa-rm@lists.oasis-open.org
Subject: Re: [soa-rm] Types of Services (RE: [soa-rm] Definition of "Service Consumer")

A dll is just an object library. It may look like this:


UUIDGenerator {
generateUUID():string
}

That looks like a service to me. You see, it doesn't matter how services are offered, as long as they are offered in a way that is appropriate for their environment. Some people will build SOAs that use Java classes and rely on rmiregistry for advertisement, whilst others will use UDDI and WSDL/SOAP.

An SOA is an SOA regardless of the physical boundaries it does or does not cross.

-Matt

On 11-Apr-05, at 9:04 AM, Christopher Bashioum wrote:


But I thought we agreed earlier on that advertisement was an important part of the SOA model - how would a dll advertise?  Through the registry of a given OS, but is it available to handle requests from outside the given OS?  (I'm obviously not an expert on dlls ; )
 
Intuitively, I am bothered by the idea that a dll can be a service within an SOA.  It is more of a gut thing.  Since there is no definitive SOA RM (that's what we are in the process of creating), I have no basis for my feeling other than that all of the SOA implementations I have looked at span process and machine boundaries.
 
I thought Microsoft considered the .net environment to be SOA, but I have not heard them talk about dlls as SOA.  Again, I am no expert in .net, but I thought .net got into the cross-machine boundaries, wherease .dlls are limited to the given OS.


From: Matthew MacKenzie [mailto:mattm@adobe.com]
Sent: Monday, April 11, 2005 11:47 AM
To: Christopher Bashioum
Cc: soa-rm@lists.oasis-open.org
Subject: Re: [soa-rm] Types of Services (RE: [soa-rm] Definition of "Service Consumer")

A .so or .dll may actually be a service within an SOA, yes.

I am balking at the idea that a service requires a request bearing a message for either input or output. I am worried that by doing this we are just defining WSDL without an XML binding :-)

-Matt

On 11-Apr-05, at 8:36 AM, Christopher Bashioum wrote:


But doesn't this leave open an SOA being just a form of OO?  Would a C++ library be an SOA?  Could you elaborate on your use case- maybe that will help me understand better.  In particular, i'm wondering about the "automatically perform" portion.  This may be a difficult area, because we don't want to get too concrete, but at the same time we shouldn't try to be so abstract that anything goes. 
 
Maybe we can put in a placeholder for the message or bus mechanism, and see if it can drop out of the RM later or if it begins to show itself as an essential part of the RM (at which time we can decide what to call it).


From: Matthew MacKenzie [mailto:mattm@adobe.com]
Sent: Monday, April 11, 2005 11:27 AM
To: Christopher Bashioum
Cc: soa-rm@lists.oasis-open.org
Subject: Re: [soa-rm] Types of Services (RE: [soa-rm] Definition of "Service Consumer")

A service is merely something that performs some task(s) on behalf of / for a consumer. Defining the mechanics of how the service knows what or when to do is, to me, not in scope for SOA-RM.

I envision services residing within an SOA that automatically perform services on behalf of a dynamic set of consumers based on their own determination of the requirement for a service to be rendered. I want this kind of use case to be recognized within the reference model without needing to explicitly spell it out.

-Matt


On 11-Apr-05, at 7:27 AM, Christopher Bashioum wrote:


Ken and Joe,
 
do all services have a 2-way communication mechanism, or is it possible to have a service that just consumes messages (or just sends messages)?  The reason I am asking, is that it looks like all the interactions mentioned so far involve a request and a response, but I am wondering about the idea of events or broadcasts. 
 
 


From: Ken Laskey [mailto:klaskey@mitre.org]
Sent: Monday, April 11, 2005 12:34 AM
To: Chiusano Joseph
Cc: Christopher Bashioum; soa-rm@lists.oasis-open.org
Subject: Re: [soa-rm] Types of Services (RE: [soa-rm] Definition of "Service Consumer")

Let me suggest the following:

A data resource is a source of content. It accepts a request and returns a value or set of values in response. The return can be an entity (such as a particular schema), an attribute of an entity (such as when the schema was last modified), or any numerical or textual value or set of values. The content can be static objects stored in some repository or dynamically generated through the use of a processing resource. Data about a missile that is stored in a database is content. The weather forecast for tomorrow is content generated from a weather simulation. In a net-centric environment, the requester does not know the format from which the response is retrieved or how it is generated.

A processing resource is one that accepts a task and return a status indicating the extent to which the task was completed and information on how the state of entities changed as a result of the processing. One or more processing resources may be invoked as part of a process of submitting a query and being returned a response. From the standpoint of a user (either human or machine), it is unimportant what combination of data and processing resources are invoked as long as the request is satisfied.

Services interact with (i.e. use, invoke, access, ...) these resources.

Ken

On Apr 10, 2005, at 2:00 PM, Chiusano Joseph wrote:


I wonder if the roles a service can play - or, perhaps one can say, the
general types of services that can exist - have any bearing on our RM at
all, in an indirect way.

Put in simple terms, one may say that there are - in general - 3
overarching "types" of services. These correspond to 3 of the layers of
the general "integration stack" (data, application, and process):

(1) Data-Oriented Service: Primary role is to accept and process data,
or provide data based upon a request.

Two general types:

(a) Data Processor*: Accepts as input a set of data, processes that
data, and (optionally) sends a response. The response may simply be an
acknowledgement, or another set of data to be processed by the service
requester**.

Ex: Simple form acceptance service, such as a loan application form
service acting on behalf of multiple banks (routes to proper bank and
sends back acknowledgement to form submitter)

(b) Data Provider: Provides streaming data, or a set of data upon
request.

Ex's: RSS news feed (streaming data), stock quote (set of data upon
request - given stock ticker symbol)

*need better term - using this for illustration purposes only
**using term "requester" for now since we have not established our
perferred term

(2) Application-Oriented Service (aka "Function-Oriented Service"):
Primary role is to accept a command and carry out processing based on
that command, in a singular fashion (i.e. does not invoke other
services).

Ex's: Inventory verification service (accepts item #, responds with
whether or not it is in inventory), shipment cost calculation service

(3) Process-Oriented Service: Similar to Application-Oriented Service,
but invokes other services in carrying out its processing (i.e. it
embodies the definition of an overarching process).

Ex: Order processing service (checks customer credit, checks inventory,
does shipment cost calculation, etc.)

Thoughts?

Joe

Joseph Chiusano
Booz Allen Hamilton
Visit us online@ http://www.boozallen.com



-----Original Message-----
From: Christopher Bashioum [mailto:cbashioum@mitre.org]
Sent: Thursday, April 07, 2005 12:49 PM
To: soa-rm@lists.oasis-open.org
Subject: RE: [soa-rm] Definition of "Service Consumer"

When we talk about service consumer vs. provider in this
sense, I think we need to separate the "static" entity from
the dynamic role that said entity plays. A given entity can
be both service provider (in which case it publishes it's
service description) and service consumer (in which case it
binds to another service provider in order to accomplish its
own service).

So...to re-word your statement a little: An entity that binds
with a service is playing the role of service consumer.

-----Original Message-----
From: Vikas Deolaliker [mailto:vikas@sonoasystems.com]
Sent: Thursday, April 07, 2005 12:21 PM
To: 'Frank McCabe'; soa-rm@lists.oasis-open.org
Subject: RE: [soa-rm] Definition of "Service Consumer"


Using the publish/find/bind framework of SOA...

The entity that publishes is certainly not the consumer. The
entity that
finds may or may not be the consumer but the entity that
binds is certainly
the consumer.

So an entity that "binds" with a service would be the closest
to a service
consumer.

Vikas

-----Original Message-----
From: Frank McCabe [mailto:frank.mccabe@us.fujitsu.com]
Sent: Thursday, April 07, 2005 9:00 AM
To: soa-rm@lists.oasis-open.org
Subject: Re: [soa-rm] Definition of "Service Consumer"

There is a distinction between the software *entity*
(agent/component/J2EE bean/.../) that interacts with a
service in order
to achieve some goal, and the person or persons for whom that
interaction is taking place.

The reason that this distinction is important is similar to the
distinction between a service interface and the service itself:
accessing your bank account from an ATM or on-line will use different
interfaces but ultimately all use the same service.

Here is an example of why its important: the appropriate
business logic
to apply to a service request will depend on many factors:
the means by
which the request was delivered, the request itself and the
person (or
persons) for whom the request was made. This last aspect is
completely
independent of mode of requesting and is purely business/application
specific.

Incidentally, the above definition: "an agent that interacts with a
service in order to achieve a goal" seems to be a reasonable
definition
of a service requester.


On Apr 7, 2005, at 7:23 AM, Gregory A. Kohring wrote:


Matthew,

OK, here a fewer other choices which might be deemed more
"respectful"...

Service Consumer:

1) End-user of a service.

2) An agent which, acting on behalf of its owner, uses a service.

3) An entity which utilizes a service

4) An entity which consumes the product or information produced by a
service.


Note all of these definitions depend upon the definition of the
term "service". Have we agreed on this already? Perhaps we should
start there first...


-- Greg



Matthew MacKenzie wrote:

I think services deserve respect, lets try not to exploit them :-)
Gregory A. Kohring wrote:

Thomas,

Perhaps one should use a somewhat broader definition
which captures

the human user as well:

Service Consumer: An entity which exploits a service.


-- Greg


Thomas Erl wrote:


Now that we've decided on the term "service consumer" it may be
useful to formally define it. The term "consumer" is used by the
WS-I Basic Profile wherein it is simply defined as
"Software that

invokes an instance."

Thomas





--

======================================================================

G.A. Kohring
C&C Research Laboratories, NEC Europe Ltd.

======================================================================









------------------------------------------------------------------------------------------
Ken Laskey
MITRE Corporation, M/S H305 phone: 703-883-7934
7515 Colshire Drive fax: 703-883-1379
McLean VA 22102-7508


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