normative keywords (was Re: FYI : a new draft of "QA Framework - Specifications Guidelines" published)

Alex, thanks for your careful reading and comments, and thanks Dom for your 
extensive replies.

Note.  I'm going to try to break out my comments into some separate 
threads.  The first is...

About normative keywords usage:

At 04:09 PM 9/3/02 -0400, Dominique Hazaël-Massieux wrote:

>[sorry for the rather long mail]
>
>Le ven 30/08/2002 à 17:06, Alex Rousskov a écrit :
>[...]
>
> >       "1.4 Terminology The keywords "MUST", "MUST NOT", ... will be
> >       used as defined in RFC 2119"
> >
> > The document does not use MUST, SHOULD, or MAY keywords. Or, at least,
> > these words are not used in the way suggested by RFC 2119 (capitalized
> > to make them look like keywords).
>
>True, I think we'll need to clarify this indeed (we have 3 non
>capitalized shall, 30 should, 16 required, 37 must, 7 optional, 5
>recommended and 73 may according to [1]).

I agree that we have some work to do here.  There are a couple of aspects:

a.) as Alex points out, there is no capitalization in the text.  So we 
either need to do that, or say in 1.4 that use in the text is not 
capitalized (contra the 2119 recommendation).  It is probably a good idea 
to capitalize, because it then you may write normal non-normative prose 
without having to do verbal gymnastics to avoid them.

b.) #a is pretty much editorial, but before we could for example capitalize 
in the text, we need to decide which level of normative-ness applies in 
each situation.   This is open Issue 64 [2] (note, it had been marked as 
closed due to some editor's error, but I have fixed that).  There is a 
similar open issue [1] for Operational Guidelines.

c.) In preparation of the most recent SpecGL draft, it seemed to me that we 
could probably limit all 2119 normative keyword usage (i.e, in the 
normative 2119 sense) to language within checkpoints.  Then, within the 
checkpoint, the subject of such a keyword becomes a MUST requirement for 
satisfying the checkpoint, and there could be SHOULD and MAY provisions as 
well.  I.e., the use of "imperative language" within the text of the 
keyword need not be strict.  This may provide the flexibility to deal with 
some of the other issues that Alex raises, about subjectivity within some 
of the checkpoints.

Regards,
-Lofton.

[1] http://www.w3.org/QA/WG/qawg-issues-html.html#x39
[2] http://www.w3.org/QA/WG/qawg-issues-html.html#x64

Received on Tuesday, 3 September 2002 14:47:32 UTC