Re: Part I of Response to Peter F. Patel-Schneider

You are correct.

We should be uniformly using IRI, which is absolute.  I had mistakenly
thought that we should be using absolute IRI, which, as you say, does
not include a fragment.

Changes to be made:
Syntax: absolute IRI -> IRI (twice)
	IRI references -> IRIs (twice)
RDF Mapping: IRI reference -> IRI (about 9 times)
Manchester Syntax: absolute IRI -> IRI (twice)

peter


From: "Michael Schneider" <schneid@fzi.de>
Subject: RE: Part I of Response to Peter F. Patel-Schneider
Date: Sun, 5 Apr 2009 12:34:03 +0200

> Peter F. Patel-Schneider wrote on April 02, 2009:
> 
>>>>Terminology change:
>>>>
>>>>"IRI reference" -> "absolute IRI"
>>>
>>> The (consistent) use of "IRI reference" in the document was
>>deliberate,
>>> because the term "URI reference" is (also consistently) used in the
>>> original RDF Semantics document. In general, I wanted to avoid
>>> terminological deviation from the RDF Semantics. I also want to note
>>> that the term "IRI reference" is used in the IRI specification itself
>>> (RFC 3987).
>>>
>>> Nevertheless, I would agree to change the term, if "IRI reference"
>>would
>>> not be in use in the rest of the OWL 2 document suite. However, I can
>>> see that this term is frequently used in at least the Structural
>>> Specification and in the RDF Mapping.
>>>
>>> As a consequence, I would prefer not to change the current use of "IRI
>>> reference".
>>
>>This is not a request for a wording change just for stylistic reasons.
>>My belief is that IRI reference is technically incorrect, as it includes
>>relative IRIs.  SS&FS has already made this change.  RDF uses URI
>>reference to mean absolute URI with optional fragment.
> 
> Ok, "absolute resource identifier with optional fragment" is what I want to
> refer to, either, because all our built-in vocabulary terms are composed
> with a fragment "#foo" attached.
> 
> So I had a look in RFC 3987:
> 
>  <http://www.ietf.org/rfc/rfc3987.txt>
> 
> According to the BNF in Section 2.2 (and hoping that the non-terminals stand
> for what they are called), an "IRI reference" can be either an "IRI" or a
> "relative reference":
> 
>   IRI-reference = IRI / irelative-ref
> 
> So you seem to be right. In this case, it is at least not correct to use the
> term "IRI reference" in Section 2.1, where it is said that the nodes of
> triples may be "IRI references". 
> 
> Now, looking further to the BNF, in order to see what is the correct term
> for referring to an "absolute resource identifier with optional fragment",
> there is
> 
>   IRI = scheme ":" ihier-part [ "?" iquery ] [ "#" ifragment ]
> 
> and, AFAICT, this has the form I am looking for.
>  
> In your original mail, you suggested "absolute IRI", but the BNF tells:
> 
>   absolute-IRI   = scheme ":" ihier-part [ "?" iquery ]
> 
> i.e. the optional fragment is missing.
> 
> So the winner seems to be "IRI".
> 
> If you agree, I will replace /every/ occurrence of "IRI reference" by "IRI"
> in the RDF-Based Semantics.
> 
> In addition, I would then suggest to use "IRI" consistently everywhere in
> our documents (I believe that we never talk about relative references, at
> least not in the core documents (perhaps in OWL/XML, I don't know)). There
> are still many occurrences of "IRI reference" in the Mapping, and at least
> two in the Structural Specification.
> 
> Do you agree with this approach?
> 
> Michael

Received on Sunday, 5 April 2009 16:51:57 UTC