oddities in our Conformance Model

Here [1] in our Conformance Clause, there is something that I find confusing:

[1] http://www.w3.org/QA/Group/2005/02/qaframe-spec/#normative-parts

We say that Requirements are normative, and everything else including Good 
Practices is informative.  I find that odd.  In my view, GPs are normative 
but optional, like a SHOULD.  The fact that GPs appear in our ICS, and the 
fact that the language and wording of GPs is exactly identical to that used 
in Rqts -- these reinforce the normative/optional view.

The question is complicated by the fact that we don't have a definition of 
normative, neither in SpecGL nor in the "comprehensive QA Glossary".  We 
have to go back 18 months [2] to find a definition of normative (at least 
one which is the result of QAWG deliberation and consensus):

>normative text
>     text in a specification which is prescriptive or contains conformance 
> requirements.

[2] http://www.w3.org/TR/2003/CR-qaframe-spec-20031110/definitions#definitions

To me, the GPs fit the definition -- you can conform to the GPs.  I could 
as easily imagine writing test cases for the GPs as for the Rqts.  That 
would seem odd for "informative" stuff.

In [1], we sort of sweep the question under the rug by saying:
>Text designated as normative is directly applicable to achieving 
>conformance to this document.  Informative parts of this document consist 
>of examples, extended explanations, and other matter that contains 
>information that should be understood for proper implementation of this 
>document.

IMO, the difficulty arises because we want a simple conformance model with 
a single conformance designation:  "Conforming" (satisfies all Rqts).  The 
problem could be solved by some additional designation such as "Conforming 
PLUS", but we avoided such complication (rightly, I believe).  It could 
also be solved by classifying the GPs as "normative, optional", saying that 
it's better than plain Conforming to satisfy as many GPs as possible, but 
not defining any designation other than "Conforming" ( == "does all Rqts").

We should eat our own dogfood -- clear definitions and clear conformance 
model.  In my view, we don't do that now.  I'm not suggesting that we 
should fundamentally alter what it means to be SpecGL-Conforming, but that 
we ought to clean up how it's structured and presented.

Regards,
-Lofton.