Re: follow-up discussions on Boston F2F, 4 Friday 2005 AM

At 04:16 PM 3/14/2005 +0100, Dominique Hazaël-Massieux wrote:
>Le lundi 14 mars 2005 à 08:06 -0700, Lofton Henderson a écrit :
>[...]
> > >It was decided to make it normative since we decided to require specGL
> > >implementors to fill it up to claim conformance to SpecGL.
> >
> > I don't think it is normative.  It does not fit our definition of
> > normative, "prescriptive or containing conformance requirements".
>
>Well, I think is becomes prescriptive once you start requiring it to
>claim conformance;

What behavior, requirement, or characteristic does it prescribe, and on whom?

>but feel free to ask the Chair to re-open the issue.

Chair:  please re-open.

>(my gut feeling with normative vs informative is: if it's normative, you
>can't remove it from the spec without changing the way you conform to
>it; in this case, removing the ICS would change the way you conform to
>SpecGL)
>
> > (Btw, "normative" has disappeared from SpecGL Glossary, from the
> > "all-inclusive QA Glossary", etc.  When did that happen (LC?)?  Given the
> > amount of time we spent arguing the definition, around Last Calll time, it
> > would seem useful to keep the resolved definition in the Glossary.
>
>Agreed; could you raise a separate issue about this so we don't lose
>track of it.

+Issue,
+Agenda (to ensure the issue is added).


> > > > As defined Thursday (Boston), isn't the ICS an (unsubstantiated) 
> claim of
> > > > spec authors' intent to conform to SpecGL?
> > >
> > >... until it is required to claim conformance.
> >
> > It is still unsubstantiated and unsupported (according to Boston) -- it is
> > simply a tool for making a claim.
>
>You don't distinguish nature and function; in its nature, an ICS is only
>a tool; but when you require someone to use the tool, it does become (in
>its function) a substantive part.

That raises an interesting question...   do we now require SpecGL 
implementors to use exactly the tool we provide, exactly our ICS?  Earlier 
SpecGL versions, we granted leeway about what ICS to use -- use the one 
provided by the spec, or another one (e.g., a testing service might draft 
its own, better, more comprehensive ICS.)

IMO, our original orientation was correct -- use some ICS, and ours is a 
suitable example, in case you don't want to make your own.

-Lofton.

Received on Monday, 14 March 2005 15:40:40 UTC