- From: David Booth <david@dbooth.org>
- Date: Fri, 13 Dec 2013 22:12:22 -0500
- To: Richard Cyganiak <richard@cyganiak.de>
- CC: Guus Schreiber <guus.schreiber@vu.nl>, public-rdf-comments <public-rdf-comments@w3.org>
Hi Richard, On 12/13/2013 07:30 PM, Richard Cyganiak wrote: > David, I don't understand your point. It seems fundamentally misguided > to me. > > RDF Concepts is a specification. It defines a data model. > > Why would we say, to pick an arbitrary other definition from the > document: “A triple is *intended* to consist of subject, predicate and > object. In cases where it does not, someone might run into trouble and > disappointment.” We wouldn't say that because a triple consists of subject, predicate and object **by definition**. But the definition of RDF does *not* restrict its use to only one interpretation. The RDF spec does not in any way prevent one interpretation from being applied to one graph, and another interpretation from being applied to a different graph. > > Obviously people can violate a specification, and of course it happens > all the time, sometimes intentionally but more often unintentionally. In > the wild, you can find stuff that doesn't conform. No one with a minimum > of common sense can be surprised by that. > > You are calling a *specification* “misleading and wrong” on the basis > that people can use the specified technology in ways that don't fall > inside the spec? No, the use of different interpretations does *not* fall outside of the spec. The spec is completely agnostic about what interpretations are used. The use of one interpretation with one graph, and another interpretation with another graph, does not in any way violate the RDF spec. > > Two more points inline. > > On 13 Dec 2013, at 23:10, David Booth <david@dbooth.org > <mailto:david@dbooth.org>> wrote: > >> Hi Richard, >> >> On 12/13/2013 04:51 PM, Richard Cyganiak wrote: >>> Hi David, >>> >>> On 13 Dec 2013, at 19:23, David Booth <david@dbooth.org >>> <mailto:david@dbooth.org>> wrote: >>>> No, I cannot live with this. The current draft of the RDF Concepts >>>> says: >>>> >>>> "IRIs have global scope: Two different appearances of an IRI denote >>>> the same resource. >>>> >>>> and that is simply misleading and false, as explained here: >>>> http://lists.w3.org/Archives/Public/public-rdf-comments/2013Dec/0073.html >>> >>>> >>> PLEASE, PLEASE, PLEASE don’t do this. >>> >>> It is true that an IRI can denote different things in different >>> interpretations. But we are talking here about a brief and >>> *informative* introduction to semantic web architecture of two pages. >>> It cannot, and should not, get into the business of explaining >>> interpretations and possible worlds. >> >> I agree with that. But that doesn't mean that it needs to be >> misleading and wrong. A balance is possible. >> >>> >>> So we are concerned here with only one possible world, the one we >>> live in. >> >> I do not agree. I do not believe that that is in fact how we use RDF. >> >>> In this particular possible world, an IRI denotes the same >>> thing wherever it occurs. For the vast majority of readers, this is >>> all they ever need to know. >> >> I do not agree. I think we need to alert them to the fact that, >> although this is the *goal* and the *intent*, it is *not* the reality. >> And they need to watch out for this when merging graphs, or they're >> going to run into trouble and disappointment. >> >>> >>> Now, you are right, RDF Semantics introduces the notion of >>> interpretations, and an IRI can denote different resources in >>> different interpretations. But this denotation of different resources >>> is not even a *feature*. It is simply part of the formalism that >>> happens to be used to define what entailments are correct. Had the >>> semantics been formally defined using inference rules rather than >>> model theory, then the phrase you quote would be absolutely correct. >> >> I can't comment on that, as I don't know exactly how it would look. >> >>> >>> In summary, David, you give us two alternatives. >>> >>> a) Either we need to introduce a brief informative account of the way >>> IRIs work on the semantic web with caveats about multiple possible >>> worlds. b) Or we can’t tell people that an IRI that occurs twice >>> should always be taken as identifying the same thing. >> >> But those are not the only two options. I think we can easily say >> that different occurrences of an IRI are *intended* to denote the same >> thing. That would prevent the statement from being *wrong*. And we >> can give a pointer to a more complete explanation, > > To quote the paragraph that immediately precedes the offending bullet point: > > [[ > Guidelines for determining the referent > <http://www.w3.org/TR/rdf11-concepts/#dfn-referent>of anIRI > <http://www.w3.org/TR/rdf11-concepts/#dfn-iri>are provided in other > documents, like/Architecture of the World Wide Web, Volume One > <http://www.w3.org/TR/webarch/>/[WEBARCH > <http://www.w3.org/TR/rdf11-concepts/#bib-WEBARCH>] and/Cool URIs for > the Semantic Web <http://www.w3.org/TR/cooluris/>/[COOLURIS > <http://www.w3.org/TR/rdf11-concepts/#bib-COOLURIS>]. A very brief, > informal and partial account follows: > ]] > > Where WEBARCH contains an entire section on URI collisions on the cost > thereof. And isn't the warning that what follows is a “very brief, > informal and partial account” enough? I think that part above is an excellent addition, but no, I don't think it is enough. In fact, even the way it is phrased above is somewhat subtly misleading, because it says "Guidelines for determining *the* referent" (my emphasis), thus wrongly implying that an IRI has only *one* referent. I don't think that phrasing by itself is harmful enough to need changing. But to later *explicitly* claim that "IRIs have global scope: Two different appearances of an IRI denote the same resource" seems too harmfully misleading to me. > >> perhaps saying that if an IRI either accidentally or otherwise was >> used to denote different resources in two source graphs, then this may >> cause problems when merging those graphs, > > But WEBARCH says this (in a non-RDF-specific way)! Yes, exactly. The problem is that the offending bullet denies it. > >> because RDF *assumes* that the IRI denotes the same resource >> everywhere it appears in the merged graph. I think it is important to >> alert readers to this real life difficulty with graph merging, rather >> than leading them down a garden path, only to be confused and >> disappointed when graphs do *not* merge as expected. >> >> My original comment included a very modest suggestion for fixing this >> text: >> http://lists.w3.org/Archives/Public/public-rdf-comments/2013Oct/0008.html >> [[ >> "IRIs are *intended* to have global scope: Two different >> appearances of an IRI are *intended* to denote the same resource." > > Whose intentions are we talking about now? What is this passive voice > business? The intent of the Web architecture, and the intent behind RDF. > What if someone publishes two occurrences of an IRI with an > *intent* that they denote different resources? I believe *you* linked to > exactly that kind of example earlier. The sentence is now less clear and > by your own yardstick just as wrong! Well, I guess "By design" could be added at the beginning of the sentence, like in AWWW http://www.w3.org/TR/webarch/#id-resources "By design, IRIs are *intended* to have global scope: Two different appearances of an IRI are *intended* to denote the same resource." Would that be clearer? David > > Richard > > >> (However, the RDF Semantics explains how an IRI may denote >> different resources in different interpretations.) >> ]] >> >> I am sure that text could be improved -- and my suggestion above would >> improve it to avoid mentioning interpretations altogether -- but the >> point is that this text does not need to be complicated or too highly >> technical to avoid being misleading and wrong. >> >>> >>> Either option is harmful to the intended audience of RDF Concepts. >>> All in the name of being *technically* correct. >> >> No, not only in the name of being *technically* correct, but also in >> the name of not being *misleading*. >> >> David >> >>> >>> Again, it’s a *non-normative* section, it’s the *introduction*, it’s >>> intended to be understandable by people who will never look at RDF >>> Semantics, and the sentence is even true within any given single >>> interpretation! >>> >>> Best, Richard >>> >>> >>
Received on Saturday, 14 December 2013 03:12:51 UTC