RE: Proposed further response regarding CoP

David,

 

The content looks good.  I would add two sentences at the very end
saying "In summary, the QA Working Group feels that the term "class of
products" is a useful and well-defined concept in our Specification
Guidelines.  The Working Group would be more than happy to consider
specific wording changes that you suggest to make the "class of
products" concept clearer and more precise in our Guidelines."

 

Mark

 

 

 

****************************************************************
Mark Skall
Chief, Software Diagnostics and Conformance Testing Division
Information Technology Laboratory
National Institute of Standards and Technology (NIST)
100 Bureau Drive, Stop 8970
Gaithersburg, MD 20899-8970

Voice: 301-975-3262 
Fax:   301-590-9174 
Email: skall@nist.gov
****************************************************************

-----Original Message-----
From: www-qa-wg-request@w3.org [mailto:www-qa-wg-request@w3.org] On
Behalf Of david_marston@us.ibm.com
Sent: Friday, May 27, 2005 3:16 PM
To: www-qa-wg@w3.org
Subject: Proposed further response regarding CoP

 


Context: 
Tracked as issue 1052: 
http://www.w3.org/Bugs/Public/show_bug.cgi?id=1052 
Initial WG attempt to dispose of the issue was sent 3 May 2005 [1] 

Not all specs have a CoP dilemma. Those that are defining real
abstractions may not directly impact vendor products. I have recently
suggested [2] some improvements to specification categories for ViS.
There are also non-specification pseudo-specs, which issue 1061 [3]
dealt with. 

Here is a propsed response to Paul Grosso's 25 May missive [4]. Mark
Skall: please scrutinize this and amend. 
=========================================== 
Please understand that the Specification Guidelines are a part of a
Quality Assurance practice that has been suggested for W3C Working
Groups. Significantly, another part of the practice is for WGs to issue
test materials as well as documents. When a WG contemplates assembling
and issuing a set of test cases, they must consider which class of
products will be the test subjects. The test cases will be applied to
each of the subject products, all of one class and supposedly
interoperable, and their respective results will be compared against the
results that represent the standard of conformance. The notion of
measuring conformance through testing motivates the SpecGL requirement
for a "conformance clause" [A] of each spec, so that vendors intending
to build conformant products can determine how to conform and which of
their products could be subject to application of a W3C-sanctioned
testing regime. 

Your objection says: 
>Furthermore, many of the core XML specs are, in fact, referenced 
>by other specs that may well be applicable to other classes of 
>products not mentioned in the core spec. 

Each spec must make its own designations of the class(es) of products
for which it intends to define conformant behavior, if any. If it is
also cited as normative by additional specs, each of the other specs is
designating their own class(es) of products for which they intend to
define conformant behavior, and so on. A very abstract specification
such as InfoSet may not define conformant behavior of any class of
product, but still be available to be cited normatively by other specs.
In this paragraph, the word "designate" means that the WG commits to
providing an objective standard by which an independent test lab can
measure conformance of individual products (instances of the class of
products). Such a commitment is fulfilled by appropriate wording in the
spec (e.g., "A conforming XML processor MUST [exhibit certain
behavior]...") and issuance of test cases will extend the WG's
fulfillment. 

In addition to providing an objective standard of conformance and
possibly some conformance tests for one or a few classes of products,
the spec may have the effect of imposing constraints on, and giving
guidance to, developers of products in other classes. The WG may wish to
recognize in the spec that the spec has such an impact, while stating
that they do not intend to provide an objective standard of conformance.
When the WG examines the full range of products that can be impacted by
their spec, they can ask themselves, for each class of product: Do we
intend to publish conformance measurement criteria and tests? The
class(es) of product for which they answer "yes" are the one(s) that
must be itemized in the conformance clause. (If this analysis proceeded
product-by-product instead of class-by-class, it would be an exercise in
discriminating against certain vendors.) A class of products for which
the WG answers "yes" is one where the WG takes direct action to impose
criteria for measurable interoperability. 

[A]
http://www.w3.org/TR/2005/WD-qaframe-spec-20050428/#include-conformance-
clause-principle 
=========================================== 
I hope that the above has sufficient manifesto-like quality to not
merely assuage the skeptics, but get them to see that they are
constraining their own workload if they adopt our approach. 
.................David Marston 

[1] http://lists.w3.org/Archives/Public/www-qa/2005May/0041.html 
[2]
http://lists.w3.org/Archives/Public/www-qa-wg/2005May/att-0016/00-part 
[3] http://lists.w3.org/Archives/Public/www-qa/2005May/0015.html 
[4] http://lists.w3.org/Archives/Public/www-qa/2005May/0112.html