W3C home > Mailing lists > Public > www-ws-arch@w3.org > March 2002

RE: D-AG0008:

From: Timothy N. Jones <tim@crossweave.com>
Date: Mon, 25 Mar 2002 11:00:22 -0800 (PST)
Message-ID: <10033531.1017082822583.JavaMail.tomcat@linux>
To: "Austin, Daniel" <Austin.D@ic.grainger.com>, www-ws-arch@w3.org

> Hi Tim,
> 	Just a few comments on your write-up on goal 8. I don't believe that
> we should use the words "complete" or "self-contained". I don't think that
> either of these applies to an achitecture in general, and especially not
> one that intends to apply to an unknown problem domain such as Web
> We are not in a position to be able to determine completeness, and in fact
> believe that we cannot and should not attempt to design the architecture
> with this idea in mind, but that we should explicitly design it to be an
> open architecture that provides sufficient flexibility to allow for the
> addition of new technologies and services as yet unknown.

The terms I chose for this could be misleading, and I would welcome
suggestions for better ones.  I too think that the architecture should be
open and extensible with future technologies.  My point was that all
technologies, once identified as relevant to web services infrastructure,
should have a place in the architecture.  For example, if transactions
requires reliable messaging, and we agree that transactions has a place in
the architecture (i.e., we can draw it on "the diagram"), then reliable
messaging needs to have a place in the architecture too, else we have a
hole.  The architecture may need to evolve if new types of technolgies (not
new instances, but new types) come along in the future, but I believe that
it should encompass the union of all common/foundation/infrastructure web
service related technolgies at any given time. 

> 	One issue regarding this goal: this goal talks about the definition
> of the architecture, i.e. the reference architecture that we define. This
> reference architecture should be consistent and coherent. This applies to
> both the reference architecture itself and the document that contains its
> definition. 

> 	To be honest, I don't understand your defintiions of consistency and
> coherence. Further, I don't see what OOP design patterns or simplicity
> to do with either, as these issues seem to me to be orthogonal. Here is a
> loosely worded definition of what I mean when I say "consistent and
> coherent".

I didn't actually attempt to define consistency or coherence, but perhaps
that should be the first step to define the common ground.  The OOAD metrics
I was referring to are mainly related to simplicity which I feel aids
coherency, but I may be confused.  These metrics judge a design based on
things like the number of relationships between objects and the distribution
of control in a system; there are many of these in the book by Riel [1].

> 	To be consistent means that the architecture does not do the same or
> similar things in mutually incompatible ways, that it is not
> self-contradictory. If a service is invoked to provide a stock quote using
> an XQL query in one place, it should do so everywhere throughout the
> in the same manner. There should not be wildly different means to achieve
> the same ends in the architecture.

I believe this says basically the same thing as my test of consistency
(below).  I would be fine with either wording.

> 	To be coherent means that the architectural components work together
> to form a logical whole. The dictionary does a good job with this word:
> coherent:
> a) Marked by an orderly, logical, and aesthetically consistent relation of
> parts
> b)Composed of mutually dependent parts; making a logical whole;
> as, a coherent plan, argument, or discourse.

My shot at coherency (below) was that the architecture should be compatible
with common systems-building techniques (design patterns) -- designers
shouldn't have to learn a new way of thinking to work with the architecture.
My wording was more abstract than I'd have liked, but the definition above
seems even more so and I am not sure how to write a CSF based on it that is
not similarly abstract (it seems that CSFs should be fairly concrete if they
are to have any teeth, so to speak).

> 	Hopefully this clarifies things.
> Regards,
> D- 



> > -----Original Message-----
> > From: Timothy N. Jones [mailto:tim@crossweave.com]
> > Sent: Tuesday, March 12, 2002 12:10 PM
> > To: www-ws-arch@w3.org
> > Subject: D-AG0008: "is coherent and consistent in its definition"
> > 
> > 
> > Goal 8 of the web services architecture is stated as follows:
> > 
> > 	"is coherent and consistent in its definition"
> > 
> > Following Daniel's lead, we have two questions to ask of this goal:
> > 
> > 1. Is the proposed text above sufficiently clear, concise and 
> > intelligible
> > to serve its purpose?
> > 
> > 2. What are the critical success factors that we need to 
> > achieve this goal?
> > 
> > The only issue I have regarding the wording is whether we should add
> > something to the effect that the architecture shall be "complete" or
> > "self-contained" -- when we look at the use cases for web 
> > services, any
> > technologies required to implement them should be part of the 
> > architecture.
> > 
> > I believe that simplicity is a part of coherency, but there 
> > is a separate
> > goal (D-AG0005) for that, so I don't think it needs further 
> > elaboration
> > here.
> > 
> > Regarding CSFs, I believe that a visualization in the form of a two
> > dimensional diagram of the architectural components and the 
> > relationships
> > between them is a top-level CSF.  This would preferably be a simple
> > "stack"-type picture that could be understood by a wide 
> > audience, rather
> > than something more formal such as a UML class diagram.
> > 
> > The only quantitative metrics that come to mind are from the 
> > OOAD world and
> > focus on simplicity, so I don't think they belong in this 
> > goal.  I have
> > mentioned one qualitative metric for completeness above.  
> > Other questions
> > that come to mind are:
> > 
> > "Does the architecture support the concepts used in commonly 
> > accepted design
> > patterns?" [coherency]
> > 
> > "Is there a small number (preferably one) way that a given set of
> > architecural components may be combined to achieve a particular
> > functionality." [consistency]
> > 
> > Please share any comments regarding the wording of this goal 
> > and appropriate
> > CSFs.
> > 
> > Cheers,
> > Tim
> > 
> > 
> > 
> > 
Received on Monday, 25 March 2002 14:01:02 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:40:55 UTC