Re: RDF Semantics - Identify vs. Denote distinction is not helpful

DRAFT of a letter responding to David Booth re. Issue 145. I will send this as an official response unless anyone objects in the next 24 hours. 


David, greetings. This is a response to your email, copied below, which was recorded as issue 145 by the RDF WG [1].

We do not propose to make any changes to the documents in response to your comment, for reasons which I will explain. 

The role of the RDF semantics is to give a precise account of the truth conditions on RDF graphs (more generally, all constructs which use RDF syntax.) These conditions involve, of course, the use of IRIs in RDF triples, used there as denoting names. IRIs are, however, used widely throughout the Web to informally "name" or "identify" entities in ways which pre-date their use in RDF, and which are governed or influenced by many other specifications (most notably HTTP and the various TAG publications on IRI use and meaning) and in some cases simply by accepted practices and conventions used in various Web applications or by communities of use. Their use in RDF as referring names often conforms to these widespread practices and conventions which are external to RDF, but not always. In some cases, it is necessary to refer to these 'external' mappings in order to properly define the intended meanings of RDF constructs themselves. In the current semantics document, for example, it is appealed to in section 7 when describing the semantics of datatype IRIs used in typed literals. 

The "identifies" terminology is widely use in pre-RDF Web standards and recommendations as a general term for these various kinds of 'meanings' of IRIs, motivating its use here. 

You say that this dichotomy (between identify and refer) "defeats the purpose of interpretations". We disagree. RDF interpretations are a formal device for capturing the way that IRIs refer when they are used in RDF triples and graphs. The "identify" terminology acknowledges that there are other ways that IRIs can be attached to things, defined and specified independently from RDF. Even if the world were so perfectly arranged that all these uses of IRIs agreed harmoniously in every case (this is not always true, but even if it were), it is still necessary to make the distinction, so that the RDF uses can be stated to agree with the pre-RDF usage without circularity,  as we do in section 7 of the semantics document. Indeed, the terminology to which you object was introduced in order to make it possible to state the datatype semantics in section 7 clearly and with sufficient precision. 

For these reasons, we intend to keep the terminological distinction between "identifies" and "denotes" introduced in section 4.

Please reply to indicating whether this message responds adequately to your comment.

Pat Hayes (for the RDF WG)


On Oct 1, 2013, at 8:30 PM, David Booth wrote:

> First off, I apologize for the lateness of these comments and how hastily they are written.  Given that people who are not members of the RDF working group cannot subscribe to the RDF mailing list -- even in read-only mode -- and there was no mention of it on the rdf-comments list (to which non-members can subscribe), and no mention of it in the editor's draft documents that i've been reading (in order to read the most up-to-date text), I did not realize that these documents were in Last Call.  Sorry!   I'll try to break my comments up into separately addressable issues.  Here is the first.
> In Section 4, The distinction between "identify" and "denote" does not seem helpful.  I think it adds more confusion than clarity.  AFAICT a key point of using the notion of interpretations is to allow IRIs to be mapped to entities in one's universe of discourse -- whatever real world entities one wishes to talk about.  By distinguishing between "identify" and "denote" in essence *two* mappings are being created: an identifies-mapping and a denotes-mapping.  This gives the impression that the identifies-mapping is the one that is used colloquially, but the denotes-mapping is the formal one addressed in the RDF Semantics. It seems to me that this dichotomy defeats the purpose of interpretations.  Interpretations are supposed to allow us to connect the formal semantics to the real world universe of discourse that we care about -- not to some universe of irrelevant, fictional entities that exist only in the idealized world of the RDF Semantics.
> In reading this section, I also get the impression that the motivation for this distinction is to avoid quandaries cased by having an IRI that may ambiguously denote two different things.  Defining two different notions of mapping from IRIs to resources is the *wrong* solution to that problem.  There is no justification for preferentially choosing one of those mappings over the other.  They can both perfectly well be denotes-mappings, but under different *interpretations*.  (Remember: the same IRI can perfectly well map to *different* resources in different interpretations.)  This already works perfectly under the existing RDF Semantics.
> In short, I think the definition of "identify" should be eliminated, as it adds confusion rather than helping.
> David

IHMC                                     (850)434 8903 home
40 South Alcaniz St.            (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile (preferred)

Received on Thursday, 3 October 2013 08:13:19 UTC