Re: normative and non-normative references

On 9 Mar 2009, at 14:05, Michael Schneider wrote:

> Hi Bijan!
>
>> -----Original Message-----
>> From: public-owl-wg-request@w3.org [mailto:public-owl-wg- 
>> request@w3.org]
>> On Behalf Of Bijan Parsia
>> Sent: Monday, March 09, 2009 1:05 PM
>> To: Bijan Parsia
>> Cc: Michael Schneider; Bijan Parsia; Ian Horrocks; W3C OWL Working
>> Group; Peter F. Patel-Schneider
>> Subject: Re: normative and non-normative references
>>
>> To be clear, I don't want to force the issue. For some of our other
>> documents, there are clear non-normative references. I.e., references
>> that could be dropped without affecting the reading of the spec text
>> itself or the design of the language.
>
> Ok, then my problem was with the term "normative" (as in the  
> statement: "the
> set of semantic conditions in section 5 are normative, while the
> comprehension conditions in section 7 are not".).
>
> What you are talking about here seems more of the kind that the  
> document
> /depends/ in some form on the reference. The "SROIQ paper" example  
> in your
> other mail is a good example for this, and so are many of the
> "non-normative" examples in the RDF Semantics:
>
>   <http://www.w3.org/TR/rdf-mt/#nonnormative>
>
> (The RDF Semantics refers to the OWL Reference, for example. :-))

Yep.

[snip]
> The outcome has been that, AFAICT, all current references in the  
> RDF-Based
> Semantics seem "normative" in the sense that the document depends  
> on them.

Well, that *normative* parts of the document depend on. See below.

> And I also do not plan to add links to work on which the document  
> does not
> depend.
>
> So an explicit distinction will not be necessary, and therefore I  
> prefer not
> to make it.

Michael, I see that you are strongly hostile to making this  
distinction and your analysis reflects that. I'm not sure why you are  
so hostile against it, but I would ask you to do the analysis with a  
more neutral eye.

> Analysis:
>
> * [CURIE]: Every CURIE is substitutable by a full-blown IRI, but  
> this would
> massively change the document (contains several hundred CURIEs),  
> and would
> make it virtually unreadable (if it wasn't anyway ;-)).

I can accept that as noramtive.

> * [OWL 2 Direct Semantics], [OWL 2 RDF Mapping]: These actually do  
> not occur
> in any normative context, but the (informative) correspondence  
> theorem and
> the whole discussion around it strongly depends on these documents.  
> Removing
> them would kill a whole section (at least).

I think you are reading it too strongly. This is informative as well.  
Removing it doesn't mean you can't read the spec (as in CURIEs) nor  
does it change the design of the language.

> * [OWL 1 Full]. :-) Ok, this would also kill a whole section,  
> namely that
> comparing OWL 2 Full with the old spec. And the very first  
> paragraph in the
> intro says that this document is meant to replace the old spec.

This isn't a normative section.
	http://www.w3.org/TR/2008/WD-owl2-rdf-based-semantics-20081202/ 
#Differences_to_OWL_Full

It can be removed without any change to the design or impairing your  
ability to read (esp. for the sake of implementation) the rest of the  
document.

I don't see any reference to the intro, which is informative anyway.  
I think this is, strictly speaking, informative.

I do not want to encourage readers to the old spec. It's good to have  
this section, but I shouldn't have to do a search on the document to  
find out that the reference isn't normatively referenced.

> * [RDF Concepts], [RDF Semantics]: If any ref is normative for the  
> OWL 2
> Full spec, then those (defines syntax and base semantics).

Yes. For sure.

> * [RDF Text]: I normatively refer to it in the "Vocabulary"  
> section, when
> listing the "OWL 2 Full datatypes": """The meaning of rdf:text is  
> described
> in [RDF:TEXT].""" I wonder, whether I could alternatively  
> indirectly refer
> to it by referring to the Structural Spec (I do this for the other
> datatypes). But as long as I am in doubt, I keep it in, because  
> this will
> not hurt.

This is the right way, imho.

> * [RFC 2119]: I think we all see this as a "normative" reference.
>
> * [RFC 3987] (IRI) vs. [RFC 2396] (old URI): Clearly, the newer IRI  
> spec is
> normative, and it obsoletes the old URI spec. The problem is that  
> [RFC 2396]
> was normative in the RDF Semantics, and I dedicate a normative  
> paragraph of
> the RDF Based Semantics (in section 2.1) to explicitly say that  
> [RFC 3987]
> is used instead of [RFC 2396] in the OWL 2 Full spec. So I can  
> hardly drop
> it.

These seem ok.

Cheers,
Bijan.

Received on Monday, 9 March 2009 15:23:51 UTC