Re: [SpecGL Draft] A.1 Principle: Include a conformance clause

Le 13 août 2004, à 01:49, david_marston@us.ibm.com a écrit :

>
> >What does it mean?
> >...Additionally it may specify the permissibility of extensions, 
> options...
>
> That which it doesn't specify directly, it SHOULD point to.

Are we sure that the prose of "what does it mean" and "why care" is a 
conformant prose and should use the RFC2119 keywords? The may as I read 
it here is just the english verb. If it's confusing we can change it 
and not use may. If it's the opposite we have to check the prose of 
each individual GP/Principles

>
> >Technique
> >3. If your technology is made of multiple individual recommendations,
>  >create a table of content item for Conformance, and explain that the
>  >Conformance section is in another document.
>
> I don't think it's that simple. The Conformance Clause in the "lower" 
> spec (e.g., XPath, Functions & Operators, Schema Part 2) should state 
> any expectations that it has for conformance, such as core vs. 
> optional or profiles, that would be binding against other "higher" 
> specs that use it. For example, XPath 2.0 will deprecate some 
> features, so they have defined a Backward-Compatibility Mode. They 
> require that any spec built on XPath 2.0 must specify the allowable 
> settings of that mode: always on, always off, or available to the user 
> as an option.

My point is that I do not dictate how the conformance section is 
organized as long as I can find it somewhere. It doesn't have to be 
necessary in every document, but in every document I must find an 
obvious path to the Conformance Section.

Here you are talking what should be a conformance clause. We are not 
talking about the content of the conformance clause but about the fact 
to have one or not and how to find it.


> How about this?
> 4. If your technology normatively incorporates Recommendations issued 
> by other WGs or external groups, the Conformance Clause must address 
> any variability presented in the other documents, whether by limiting 
> choices and/or providing a mechanism to allow the user to designate 
> the choice.

	content of the conformance clause... not about having a conformance 
clause

> 5. When the spec defines material to be incorporated in other 
> technologies (e.g., XPath), the Conformance Clause MUST have 
> provisions addressed to the WG that is specifying the incorporating 
> technology. At a minimum, it must state what set of features is the 
> core or minimum set (if any), and whether partial implementations are 
> acceptable.

	Same.

> (Rationale for 5: A developer of an operating system or virtual 
> machine system should be able to decide whether it's economically 
> feasible to implement the "lower" technology as a shared library, if 
> they believe that several W3C technologies will share it.)



-- 
Karl Dubost - http://www.w3.org/People/karl/
W3C Conformance Manager
*** Be Strict To Be Cool ***