Re: normative and non-normative references

From: "Michael Schneider" <schneid@fzi.de>
Subject: RE: normative and non-normative references
Date: Mon, 9 Mar 2009 11:39:34 +0100

>>-----Original Message-----
>>From: public-owl-wg-request@w3.org [mailto:public-owl-wg-request@w3.org]
>>On Behalf Of Bijan Parsia
>>Sent: Saturday, March 07, 2009 12:53 AM
>>To: Michael Schneider
>>Cc: Ian Horrocks; W3C OWL Working Group
>>Subject: Re: normative and non-normative references
>>
>>On 6 Mar 2009, at 23:27, Michael Schneider wrote:
>>
>>>> -----Original Message-----
>>>> From: public-owl-wg-request@w3.org [mailto:public-owl-wg-
>>request@w3.org
>>>> ]
>>>> On Behalf Of Ian Horrocks
>>>> Sent: Friday, March 06, 2009 11:54 PM
>>>> To: W3C OWL Working Group
>>>> Subject: normative and non-normative references
>>>>
>>>> Peter has updated Manchester Syntax to distinguish normative and non-
>>>> normative references [1].
>>>>
>>>> Can other editors please do the same.
>>>
>>> Hi!
>>>
>>> I wonder if any of the references in the RDF-Based Semantics
>>>
>>>  <http://www.w3.org/2007/OWL/wiki/RDF-Based_Semantics#References>
>>>
>>> is non-normative. What are the criteria for normative vs. non-
>>> normative
>>> references?
>>
>>If a normative criterion of your spec depends on a reference, then the
>>reference is normative.
> 
> I disagree. 

So do I, but my wording was supposed to be a reflection of W3C practice,
not the way I think things should work.

> For me, references can never be normative or non-normative
> themselves. Only the context they appear in can be. 

Agreed.

> And since I have tagged
> all non-normative parts of the RDF-Based Semantics clearly as such, I don't
> see why I should put additional (redundant!) normativity statements to the
> references list.

It's the W3C way.  (I'm not sure whether it is absolutely required, but
certainly lots of W3C RECs divide up their references this way.)

> But even if I would agree, I really don't see that such a separation is easy
> to produce in the RDF-Based Semantics. For example, you say:
> 
>>I found some mislabeled refs:
>>	[OWL 2 Structural Spec]
> 
> (Thanks for this. Will fix them!)
> 
>>But otherwise, they seem normative to me.
> 
> While Peter says in his mail:
> 
> [[
> Normative
> 
> CURIE
> OWL 2 SS&FS - because of datatypes 
> RDF Concepts - because of RDF graphs
> RDF Semantic - duh!
> RDF:Text
> RFC 2119
> RFC 3987
> 
> Non-normative
> 
> Direct Semantics !
> RDF Mapping ! - because Section 7 is informative
> OWL S&AS
> RFC 2396
> ]]
> 
> And I cannot agree with both of you. 
> 
> For example, you both say that CURIE is normative. But I don't see why. I
> could equally well use full-expanded IRIs everywhere in the document. But I
> am not sure, because the RDF-Based Semantics makes heavy use of CURIEs
> throughout the document.

You use CURIES, so the CURIE CR may be normative.  I didn't do a
complete check of everything in your document, so my characterization
might not be totally correct.

> I'm even less sure regarding OWL S&AS. Saying that OWL 1 Full is
> non-normative, as Peter does, might be valid, but only for the very
> technical fact that OWL 2 Full hasn't been spec'ed as a semantic extension
> of OWL 1 Full. 

So non-normative.  My rule of thumb is that (a portion of) a document is
normative if changing it changes your spec.  As you don't depend on OWL
1 for any technical stuff, therefore no OWL 1 documents can be normative
for you.  (I believe that this is true in general for OWL 2 documents.)

> If it had, then I would be again unclear whether the RDF
> Semantics is normative or not, since in this case it would have been fully
> covered by OWL 1 Full.

Indirectly normative, at least.  I would say that in this case if you
referenced it directly it would count as normative.

> So, there seems to be, at least, no obvious line of separation. 
> 
> Further, tagging OWL 1 Full as non-normative in the OWL 2 Full spec seems to
> me at least highly unwise, because I can imagine certain people putting this
> heavily into question. Not making an explicit distinction will save us from
> such worries.
> 
> But I also have a principle concern: I don't want to play with normativity
> statements without any necessity. I consider changing something from
> normative to non-normative to be a change in design. Not saying anything,
> again, will avoid such problems.
> 
> So is there any requirement to do this separation? Does W3C documents
> require this? Otherwise, I don't see why I should change the RDF-Based
> Semantics. I need to see a clear reason for this, and I haven't heard any so
> far. The current state is perfectly fine for me, and I don't expect anyone
> to formally object against /not/ having such a distinction. I'm not so
> certain, however, about the other way around...

I don't expect that there would be any *objections* about this at all,
just, perhaps someone politely pointing out something that they think
should be changed.

>>Cheers,
>>Bijan.
> 
> Cheers,
> Michael

peter

Received on Monday, 9 March 2009 11:11:49 UTC