- From: Phil Archer <phil@philarcher.org>
- Date: Mon, 26 Jan 2009 15:32:15 +0000
- To: Michael Schneider <schneid@fzi.de>
- CC: public-powderwg@w3.org, Stasinos Konstantopoulos <konstant@iit.demokritos.gr>
Then I must humbly apologise for our lack of manners. I'll copy and paste the comments and resolutions from the comments tracker below but the revised document [1] is probably the best guide to how we incorporated your welcome comments. I have not copied the various typos your pointed out- they were all corrected and we were grateful to you for spotting them. Again, sincere apologies for not having sent this to you a long time ago. Cheers Phil. Comment: Why the "SHOULD" here? From the assumptions stated above (wdrs:issuedBy is a subPropertyOf both dcterms: creator and foaf:maker) it follows logically that this URI denotes an instance of both classes. Resolved Yes This was a mistake as the intention of allowing either of the two Agents was wrongly encoded as subsuming issuedby under both dc:creator and foaf:maker. In the current draft using dc:Agent or foaf:Agent as filler for issuedby has been reduced to a suggestion, as issuedby is an owl:AnnotationProperty and cannot have a range restriction. Comment: Typo? "this that" Reply: No, but maybe not expressed in the clearest way possible. "It is this [new class] that is described..." This text will be removed from the next draft anyway, as this feature-at-risk has been dropped. (cf. LC-2126) Comment: Typo? Is "refer" meant? Resolved no. No, defer is the correct word, in effect the document says "don't look at this descriptor set - look at that one over there instead." You can replace this one with that one... i.e. defer to it. Comment: There is no owl:Property, only rdf:Property. Further, if you apply RDFS/OWL-Full semantics, then every instance of owl:DatatypeProperty is also an instance of rdf:Property, so the second axiom would be redundant. Reply: Fixed. Thanks. Comment: Although technically possible in OWL Full, I suggest to /not/ build custom ontology properties. AFAIK, the set of ontology properties in OWL is intended to be fixed; at least I don't see any encouragement in the OWL documents to create custom ontology properties. I suggest to make 'attribution' into an owl:AnnotationProperty, instead. The creation of custom annotation properties is ok. It is also ok to add annotation properties to instances of the class owl:Ontology. For example, it is quite common to add an rdfs:comment to an ontology. Resolved yes. Formal range restrictions have been removed from the current draft to ensure that attribution elements are owl:AnnotationProperty instances. Comment: I don't follow this argumentation. I don't see that the bNode leads to any problems when the descriptor set is defined independent on any DR. Maybe there is a misunderstanding in what the above example means. The existential variable isn't about resources from a DR, but about the values of a descriptor for such resources. The construct above tells the following: If the above descriptorset is defined for some DR, and r is a resource in that DR (i.e. if r is in the IRI-set defined by that DR), then there /exists/ some x, being an instance of class Wood, where x has a shiny finish and is made of cedar, and the following relationship holds: 'r ex:material x'. the existential variable x appears on the /right/ hand side of the relationship! If this was a misunderstanding, then I suggest to simply remove the whole paragraph. Resolved partial: Paragraph removed as a response to LC-2126, as the WG has resolved to drop this feature-at-risk. (Resolution to LC 2126 put restrictions on the semantics that can be expressed in a POWDER doc). Comment: Why a class? I think that this is a bug. PolishedCedar should probably rather be an /instance/ of class Wood. Just compare this with Example 3-10 and my comment to paragraph following Ex 3-10. The only difference between Ex-3-10 and Ex-3-11 is that the latter uses an URI where the former uses a bNode. In both cases, an instance of class ex:Wood is created. So the correct mapping should be: 1 <owl:Class rdf:nodeID="descriptorset_1"> 2 <owl:intersectionOf rdf:parseType="Collection"> 3 <owl:Restriction> 4 <owl:onProperty rdf:resource="http://example.org/vocab#material"/> 5 <owl:hasValue> 6 <ex:Wood rdf:about="http://my.example.org/myVocab#PolishedCedar"> 7 <ex:finish rdf:resource="http://example.org/vocab#shiny"/> 8 <ex:madeof>cedar</ex:madeof> 9 </ex:Wood> 10 </owl:hasValue> 11 </owl:Restriction> 12 </owl:intersectionOf> 13 </owl:Class> Reply: No, not a bug at all. The idea here was that a particular piece of PolishedCedar is the property filler, but the class of PolishedCedar pieces of woord is not defined in any available vocabulry, so the POWDER doc has to create the class and instantiate an (otherwise unknown) particular piece of wood of this class. Either way, this feature-at-risk has been removed (cf LC-2126) Comment: The OWL/XML URI is at risk: There is an ongoing discussion in the OWL WG whether there should be a URI for the XML syntax which is distinct from the OWL namespace. See <http://www.w3.org/2007/OWL/tracker/issues/109> Resolved yes: This was a mistake, left-over from a previous (unpublished) version. The relevant section (4.6) was, is fact, already using the owl2 vocabulary (http://www.w3.org/2006/12/owl2#) since the LC draft. Thanks drawing our attention to this. Comment: AFAIK, the URI <http://example.org> is reserved for such examples. And I have seen it being used elsewhere in this document. So I suggest to simply state this URI here instead of the given sentence. Resolved yes OK, but I think the existing text is clear so I've amended it to say "An arbitrary prefix used to denote an 'example vocabulary' from the example.org domain." Comment: Even though RDF/XML is the preferred RDF serialization used to represent RDF data in this document, it is probably not a good idea to talk about "RDF/XML statements" when referring to child elements of 'attribution'. I cannot see any RDF/XML-specific aspects here. These statements are probably intended to show up in every RDF serialization, such as Turtle. So I suggest to rather talk about "RDF statements". Resolved partial This had already been rephrased in response to a comment by Peter Patel-Schneider (LC-2116), that RDF statements are not about anything at all. Comment: It's misleading to talk about a "union" here, although correct in some sense. The statements are combined by "AND". I suggest to say "conjunction" instead. Resolved partial: Misleading as "union" might be, "conjunction" would be outright wrong. Will be rephrased so as to not use any logic terminology Comment: This "all resources specified by" statement is a very informal way to "enumerate" the IRIs from the iriset. I guess the reason is that the iriset can be specified by a regular expression, and so a finite enumeration is not always possible, right? I suggest to say a few words about this, because I assume that this sort of notation will confuse people. Resolved no: Please note the phrase "the IRI set has been elided for clarity (the semantics of the IRI set is discussed in Section 4 below)" immediately above Ex 3-1. We simply defer showing both POWDER/XML irisets and their RDF translations for clarity, as these have not yet been introduced, and informally describe what should be there. Please let us know if your editorial comment persist after this explanation and you think that this should be made more explicit in Ex 3-2, 3-3, and 4-3. Comment: A rational for this "MUST NOT" statement should be given. At least under RDFS/OWL-Full semantics, RDFS and OWL classes are RDF(S) resources, too. resolved yes: As you note, the restriction is there to maintain OWL-DL semantics for POWDER documents. A short note to this effected has been added to the current draft. Comment: This is a strange sentence for a spec, and it does not seem to have any relevance. I suggest to remove it Resolved yes: This is meant to discourage without prohibiting the use of RDFS properties as descriptors. Rephrased. Comment: In the axiomatic triples, hasIRI is written without a namespace prefix, but in the semantic condition below, it is written with a namespace prefix. Note that in sec. 4.3, matchesregex had also be written with the wdrs prefix. Resolved yes Thank you for spotting the mistake. Will be fixed in the next draft. Comment: In 4.3, the domain of wdrs:matchesRegex was rdfs:Resource. Now the domain of hasIRI is owl:Thing. Although, under OWL Full semantics, these two classes are equivalent, one should decide for one in order to be coherent. Resolved yes: Thanks for spotting this, the Sect 4.6 domain is wrong. [1] http://philarcher.org/powder/formal/formal-20090123-diff.html Clearly this is a temporary URI but this is the most up to date version of the document. Your suggestions are cited in the change log (section 8.2). Michael Schneider wrote: >> -----Original Message----- >> From: Phil Archer [mailto:phil@philarcher.org] >> Sent: Monday, January 26, 2009 3:03 PM >> To: Michael Schneider >> Cc: public-powderwg@w3.org; Stasinos Konstantopoulos >> Subject: Re: LC-Comment on POWDER-Semantics >> >> Michael, >> >> I'm just going through and checking that all comments that we have >> received have been answered. Yours certainly were, in that several >> changes were made [1] to the Formal Semantics document as a direct >> result of your comments, for which, I would like to again express the >> WG's thanks. >> >> What I can't find is the e-mail from us (specifically from Stasinos >> Konstatopoulos) saying "here's how we handled your comments and is this >> OK with you?" >> >> Stas - did you send such an e-mail? Michael - did you get it? > > No, I don't remember having received such an answer. > > Best regards, > Michael > >> Thanks >> >> Phil. >> >> [1] http://www.w3.org/TR/2008/WD-powder-formal-20081114/#sincelc > > -- > Dipl.-Inform. Michael Schneider > Research Scientist, Dept. Information Process Engineering (IPE) > Tel : +49-721-9654-726 > Fax : +49-721-9654-727 > Email: schneid@fzi.de > WWW : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555 > > ============================================================================ > == > > FZI Forschungszentrum Informatik an der Universität Karlsruhe > Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe > Tel.: +49-721-9654-0, Fax: +49-721-9654-959 > Stiftung des bürgerlichen Rechts > Stiftung Az: 14-0563.1 Regierungspräsidium Karlsruhe > Vorstand: Prof. Dr.-Ing. Rüdiger Dillmann, Dipl. Wi.-Ing. Michael Flor, > Prof. Dr. rer. nat. Dr. h.c. Wolffried Stucky, Prof. Dr. rer. nat. Rudi > Studer > Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus > > ============================================================================ > == -- Phil Archer w. http://philarcher.org/
Received on Monday, 26 January 2009 15:33:00 UTC