what's normative [was: Re: processing plan for SpecGL LC issues]

As I mentioned last week, the "normative" question has potential to evolve 
into a religious issue, and consume an arbitrarily large amount of 
time.  Everyone has his/her own intuitive opinion of what "normative" 
means.  Here we go...

At 08:26 PM 4/12/2003 -0400, Lynne Rosenthal wrote:
>Below is the continued processing of issues for SpecGL for Monday 14 April 
>Telecon.
>
>Topics:
>1. What is Normative - this is continued from last Thursday.  I've 
>summarized and continued the discussion.  (Issues: 36, 65, 106, 108 and [1])
>[...]
>[1] http://lists.w3.org/Archives/Public/www-qa-wg/2003Apr/0073.html
>
>+++++++++++++++++++++++++++++++++++++
>WHAT IS NORMATIVE  (36, 65, 106, 108 and [1])
>
>Summary from last week:
>- The glossary and Section 4 Definitions are Informative (#36).
>- Section 3.1 which identifies what in SpecGL is normative will be 
>rewritten  improve wording and expand bullet list as needed.  Text 
>identified by 'conformance requirements' will indicate Normative text, 
>rather than using RFC keywords (#106 and #108).  Also make sure it's 
>clear, that Normative text can be sections, paragraphs, not just a 
>sentence  thus consistent with our definition (#108).
>- We will not label Sections as normative or normative.  CONFIRM: that we 
>don't want to label Section 4 as Informative  it used to be labeled 
>(non-normative).
>
>-THINK about this (it is a prelude to the item below it) (#108)
>Is Section 1 Normative? I think it is.

I think it is not, under our current definition.  If we change our 
definition, then perhaps it is.  And the Definitions/Glossary will be, 
also, according to the criterion in your next sentence:  "This information 
is needed in order to understand and use the SpecGL."

Below, you suggest redefining "normative".  If that is the case, then last 
week's results need to be revisited (e.g., #36).

>This information is needed in order to understand and use the SpecGL.

Some of it, yes.  Other stuff, no.  For example, do you need our 
definitions of checkpoint priorities (sec 1.7) in order to understand and 
use SpecGL?  If we expand the definition of normative to encompass 
"necessary to understand and use", where does it stop?  I could claim 
(accurately!) that I need ExTech to understand and use some of our checkpoints.

>Also, to be consistent, IMO, most specs do consider this type of 
>information normative. Note, that there may be informative information 
>embedded into the section, e.g., examples.
>
>New Stuff:  (#65, #108)
>Definitions for normative and informative in Section 4.  These are more 
>narrowly focused than the definitions used in the UAAG glossary.  As per 
>[1], they contain the notion 'directly connected' to conformance  e.g., 
>test assertion or conformance requirement.  Agree?  Or should normative be 
>broader?   Do we want to change the definition?  Accept UAAG definition or 
>Volunteer to draft?

Ref:  http://lists.w3.org/Archives/Public/www-qa-wg/2003Apr/0073.html

(Contains UAAG definition, etc)

Alternatives:

1.) keep ours, explain clearly what are its limits, and what we think is 
normative in SpecGL.
2.) adopt UAAG ("..what is required for conformance")
3.) other

I favor #1 (see reference for reasons).  But I'm open to being 
convinced.  (In particular, since our definition is tighter than many 
others', maybe we will get resistance to ours?)

If anyone is going to argue for #2 ... could you circulate some 
reasons?   To me, it seems a bit vague and open ended.  (It will qualify as 
normative a superset of what would get the designation under our current 
definition.)

If we opt for #3 ... we will need a specific proposal before we can go any 
further with these issues.  In other words, "postpone" until we have a 
proposed definition to look at and discuss.

IMO, finally, it does not actually matter so much exactly how the words are 
defined, as long as it is clear in a specification (e.g., ours, SpecGL) 
exactly what is required for conformance to the specification.  For 
example, if we were to say in our "Terminology" something like, "When used 
in this specification, terms have the meanings assigned in 'Definitions' 
and 'QA Glossary'", then does it really matter whether the 
Definitions/Glossary carry the label of "normative" or 
"informative"?  (Okay, the only flaw with this observation is our CP13.2, 
"Distinguish normative and informative text.")

>
>The Checkpoint Priorities:  Are these Normative?  (#106)  (May be moot, 
>depending on resolution of definition of normative)

Question.  Does he mean the definitions in 1.7, or the priorities 
themselves that are associated with each checkpoint.

-Lofton