Re: Fwd: Re: Normative vs. non-normative references

[attempting to follow on thread from <mid:200201291358.IAA00769@tux.w3.org>]

Yes, we should try to use the term consistently.

To understand the notion of a normative reference, adopt a constraint-flow view
of the corpus of software specification literature.  Normative references
document relationship arcs over which there is material constraint flow.

If the effect of what is imported from the reference is non-binding, it is not
a normative reference.

If there is no material effect (removing the constraint doesn't include or
exclude anything from that which conforms to the current document) it should
not be recognized as a normative reference.

But otherwise, if it is the intension of the current specification to create a
pattern of practice which conforms to a constraint pattern which, when indexed
by its locus of first publication, is associated with the referenced document,
then it is a normative reference.

Al

Details and application:

At 05:53 AM 2002-01-29 , Bert Bos wrote:
>Norman Walsh writes:
>> [ This bounced the first time; retry ]
>> 
>> Hello chairs,
>> 
>> As you'll see from the thread below, Joseph Reagle suggested I forward
>> this to the chairs alias. A little background: I've been looking over
>> a bunch of "core" specifications and tracing down their normative
>> references, just to see what the set of dependencies looks like. I
>> noticed that a couple of the specs make no distinction between
>> normative and non-normative references and specifically I commented on
>> the PR comments list for...digital signatures, I think.
>> 
>> I'm in favor of the distinction myself, and propose a definition for
>> normative in the message below. Anyway, I agree with Joseph that
>> chairs is an appropriate place to talk about it. (But I'm not a chair,
>> so if there is follow-on discussion, please CC me if you want my
>> opinion :-)
>
>The CSS drafts don't always make a distinction between normative and
>informative, at least not in the formatting of the list of references,
>but there is a criterium that is often used to determine the status of
>a reference:
>
>    Can you implement this draft when you don't have access to the
>    referenced document?
>
>The primary reason we ask that is not because we want to list
>normative references separately, but because we want to be certain
>that normative references are as easily available as the spec itself.
>Thus, when the reference is to another W3C spec, we don't care very
>much whether the reference should be considered normative or not,
>because that spec is readily available; but when the reference is to,
>say, an ISO spec, we sometimes try to include the relevant definitions
>in our own spec and demote the reference to an informal one.
>
>For example, we have borrowed concepts from a Japanese standard on
>typography, but since we know that that standard is not (freely)
>available in English, we include sufficient text in our spec that you
>don't need to read the Japanese text.
>

AG::

Interesting examples.  My sense of the best practice is that in both of these
last cases the copy does not supplant the reference as normative.  It is not
grounds to downgrade the reference.  For purposes of information or
convenience, a copy is provided locally; but for purposes of definition, the
cited work rules.  If the copy and the reference clash due to blunders in
reconstructing the definition, it is in these cases an erratum and our document
should be changed, because the intent is to create practice which is
intension-equivalent to the practice adhering to the other document.

In other words, the copied content should be viewed as still dependent on the
referenced document for its initial, authoritative, definitive statement. 
This, of course, devolves into religious wars over theories of Intellectual
Property.

In the discussions of this issue in pursuit of a WAI-wide glossary, Katie
Haritos-Shea quotes the following, which is useful

<quote cite="http://lists.w3.org/Archives/Public/wai-xtech/2001Oct/0017.html">

#4

From Whatis Glossary.....................

<http://whatis.techtarget.com/definition/0,,sid9_gci214069,00.html>http://w
hatis.techtarget.com/definition/0,,sid9_gci214069,00.html

normative:

In information technology standards, normative parts of a standard are those
that specify what implementers should conform to and non-normative parts
consist of examples, extended explanations, and other matter not dealing
directly with the specifications.

<quote>

From this perspective, that 'normative' is a property of the clauses or
sub-utterances within a document, the normative references are those that build
bridges to other documents over which dependencies _from normative clauses in
the current context_ flow to their targets.

Local replication of definitive text from a source, however, should not be used
as an excuse to demote the reference, if the effect of the pattern applied both
there and here is binding here, and it is the sense of the specification that
the pattern be the same.

On where to discuss:

pro-chairs: This is where we collect those who have to apply whatever concept
we adopt.  It is where earlier threads such as how to handle dependencies on
external documents that go through incompatible upgrades have passed.

anti-chairs: The prime body of precedent in this area is in the ISO/IEC tree of
activities.  I forget the correct term for the capital-S Standards activities. 
To draw on their wisdom, we have to lure them into public discussion, hence
expose ourselves to public view, which is not a property of this list.

By the way, Susan, the thread from an entry in the manual to the discussion
that provides the rationale for the change is not to be viewed as a normative
reference.  Maybe you meant this tongue in cheek, but given the level of
confusion surrounding this notion, that is not safe in the current thread.  The
link to rationale is necessary for someone to know _why_ the change was made,
but not to know _what_ to do to obey the manual.  The decision which acts on
the rationale and authoritatively decides that the document shall change
upgrades the status, the kind of authority vested in the injunction, and the
prior discussion informs, but does not determine, that decision.  The whole
point of the document preparation process is to create a statement which is
nearly standalone; it is standalone modulo the normative references. 

If it has no bearing on recognizing what stuff obeys or does not obey the
current document, it is not normative.

This notion (of 'normative') has generated a lot of heat in the halls of the
WAI.

<http://www.w3.org/WAI/GL/Glossary/printable.html>http://www.w3.org/WAI/GL/
Glossary/printable.html

<http://www.w3.org/TR/UAAG10/glossary.html#def-normative>http://www.w3.org/
TR/UAAG10/glossary.html#def-normative

Perhaps a little light.  But also perhaps not total resolution.

<http://lists.w3.org/Archives/Public/w3c-wai-gl/2001OctDec/thread.html#101>
http://lists.w3.org/Archives/Public/w3c-wai-gl/2001OctDec/thread.html#101

Al
>
>
>Bert
>-- 
>  Bert Bos                                ( W 3 C )
<http://www.w3.org/>http://www.w3.org/
> 
<http://www.w3.org/people/bos/>http://www.w3.org/people/bos/                
              W3C/INRIA
>  bert@w3.org                             2004 Rt des Lucioles / BP 93
>  +33 (0)4 92 38 76 92            06902 Sophia Antipolis Cedex, France
>

Received on Tuesday, 29 January 2002 09:41:30 UTC