Re: definition of Class of Product

There are aspects of each definition -- Lynne's and David's -- that I 
like.  Lynne's is pleasingly brief, but I think the characteristic that 
defines the "group of products" is too imprecise.  "..that would implement 
the specification" doesn't really suggest what puts them in the same group.

I like the notion of "role" that David introduces, although it needs some 
completion -- role in what, or role related to what?  (something like 
"share a common role in conformance determinations", or "share a common 
role in interoperation scenarios of the specification's use cases," or 
...).  David's definition is, to my taste, a little too abstract.  For the 
"4. Definitions", I would like something that is more "testable", i.e., 
easier to see if a given candidate CoP in fact qualifies as a CoP or not.

So I'd prefer a hybrid (for "4. Definitions"), and maybe some of David's 
additional verbiage/examples might find a home in the new section 2.2?

-Lofton.

At 02:33 PM 4/19/03 -0400, Lynne Rosenthal wrote:
>We need to add the definition of Class of Product to the Definitions. 
>Please let me know if you have a better definition than this one.....
>
>Class of Product: the generic name for the group of products that would 
>implement the specification, i.e., target of the specification. The class 
>of product is the object of the conformance claim.


At 02:15 PM 4/20/03 -0400, david_marston@us.ibm.com wrote:
>How about something like this?
>Class of Product: a role envisioned by the spec. The spec could define 
>more than one role (e.g., client and server) while specifying the protocol 
>by which the roles interact, or it could define the behavior required of a 
>particular role (e.g., XQuery processor) and assume that other products 
>benefit when the instances of the target role (individual products) behave 
>interoperably. Continuing the XQuery example, notice that while the 
>products that fulfill the role may vary widely in their overall 
>characterization (e.g., standalone XQuery processor vs. data retrieval 
>middleware vs. various kinds of DBMS), the spec would only address their 
>behavior in fulfilling the role of an XQuery processor. In some 
>programming languages, we would say that the class of products "implements 
>an interface" so that other software can rely on a particular bundle of 
>capabilities being present.
>
>I think that the definition has to be careful to allow multiple classes of 
>product to fall into the genre of consumers of the output. In other words, 
>there can be specs that impose requirements for more than one CoP, but the 
>CoPs are all of the consumer genre. Similarly, a non-user agent might be 
>able to be both a client and a server at different times. The CoP 
>mechanism in the spec is used to define what any instance of the server 
>must do, and independently define what any instance of the client must do.
>.................David Marston