- From: Richard Cyganiak <richard@cyganiak.de>
- Date: Fri, 13 Feb 2015 18:26:26 +0000
- To: kcoyle@kcoyle.net
- Cc: Irene Polikoff <irene@topquadrant.com>, "public-data-shapes-wg@w3.org" <public-data-shapes-wg@w3.org>
Hi Karen, > On 13 Feb 2015, at 16:24, Karen Coyle <kcoyle@kcoyle.net> wrote: > > It relates thusly: if my data will stay in my closed world and will not, in that form, be used in the open world, then I can do whatever I want with it because no one else will ever see it. Heck, I don't even have to use RDF. What I do in the privacy of my own data store is my business. E.g. the closed world OWL mode. But if I think take that data and port it to the open world using those same OWL constructs, the meaning of my data in the open world is significantly different from the meaning in my closed world. Run the same operations in those two worlds and you get different results. The closed world OWL might result in inferences that you do not intend in the open world. > > That's how open world and validation relate to each other: can I use the same instance data, unaltered, in both worlds with the same meaning? If not, then I need two sets of data, one for the closed world and one for the open world. > > Not only that, there seems to be the assumption that all of the data that I will operate over has been designed for my closed world. The solutions proposed do not seem to cover the case where I aggregate data from the open world, data from some source that does not produce data to my validation specifications. That is, however, a common use case in my corner of the world. If that is NOT to be addressed here, I would like that to be explicit. This is quite incomprehensible to me. You appear to use the term “open world” to mean “an environment where multiple parties may exchange and aggregate data without prior agreement”. You appear to use the term “closed world” to mean “an environment where data is not exchanged between multiple parties, or where prior agreement about its shape and semantics exist”. But that’s not the generally accepted meaning of “open world” and “closed world”. These terms refer to two specific modes of data processing (a.k.a. reasoning), e.g., in validation and querying. Open-world reasoning is when you assume there could be additional data “out there” that you just don’t know about yet, so “missing ain’t broken”. Closed-world reasoning is when you assume that your dataset is complete, so “missing” is a validation error. I believe that you are well aware of these meanings, but choose to use these terms with a different meaning anyway. I find this highly confusing, as open-world reasoning can be quite useful in some “private/local/command-and-control” environments, and closed-world reasoning can be quite useful in “public/global/decentralised” environments. When you say things like this, it makes no sense to me: - “Data will stay in my closed world” - “Data will be used in the open world” - “Data has been designed for my closed world” - “I aggregate data from the open world” Best, Richard > > kc > >> >> Irene >> >> On Fri, Feb 13, 2015 at 10:08 AM, Karen Coyle <kcoyle@kcoyle.net >> <mailto:kcoyle@kcoyle.net>> wrote: >> >> 1) schema.org <http://schema.org> does use the concepts of domains >> and ranges, extensively, but it defines them more loosely than RDFS. >> They are fundamental to schema.org <http://schema.org> >> >> 2) are you assuming that your data is closed-world only? If it is >> not, are there implications to this use of rdfs:Class in the open world? >> >> kc >> >> >> On 2/13/15 6:35 AM, Irene Polikoff wrote: >> >> I think we agree. They don't contribute anything to validation, >> but if people want to use them that is OK. From the data >> definition/data validation perspective they will be ignored. >> >> Irene >> >> Sent from my iPhone >> >> On Feb 13, 2015, at 7:15 AM, Peter F. Patel-Schneider >> <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>> wrote: >> >> -----BEGIN PGP SIGNED MESSAGE----- >> Hash: SHA1 >> >> I don't see that this is any reason to not let people who >> want domain and >> range to use domain and range. If some people don't want >> domain and range >> then the solution for them is simple - they don't need to >> use domain and range. >> >> peter >> >> >> On 02/13/2015 02:18 AM, Irene Polikoff wrote: >> The reason to exclude domain and range is the same >> reason why Schema.org >> excluded them. They don't work in a way that is useful >> to a community >> interested in specifying what data should look like. >> >> In addition to not being useful, they also create >> problems by >> intersecting multiple ranges and domains, etc. They are >> often misused. >> >> So, one could call this RDFS- data. I don't think >> domains and ranges must >> be prohibited though, they could just be ignored. >> >> Irene >> >> On Feb 12, 2015, at 10:08 PM, Peter F. Patel-Schneider >> <pfpschneider@gmail.com >> <mailto:pfpschneider@gmail.com>> wrote: >> >> I suppose that the working group could exclude >> rdfs:domain and >> rdfs:range from the RDF graphs that it considers to be >> acceptable, just >> as OWL DL excluded certain RDF graphs. For OWL DL that >> was to achieve >> decidability and I don't see an equivalent need here. >> >> peter >> >> >> On 02/12/2015 04:03 PM, Holger Knublauch >> wrote: >> On 2/13/2015 8:19, Peter F. >> Patel-Schneider wrote: Is the working >> group producing a solution tailored for >> RDF data, where RDF >> graphs and rdf:type are important; for >> RDFS data, where >> rdfs:subClassOf, rdfs:subPropertyOf, >> rdfs:domain, and rdfs:range >> are also important; for Linked Data, >> where dereferencing and >> interlinking is important; or for >> services data, where brevity >> may be important? >> >> 2. Shapes and Classes >> >> Are shapes RDF classes, i.e., should >> shapes be the object of >> rdf:tyoe triples, participate in >> rdfs:subClassOf relationships, >> and be the object of rdfs:domain and >> rdfs:range triples? >> >> >> In both points you seem to assume that if we >> use rdfs:subClassOf >> then we also must use rdfs:domain and >> rdfs:range. Could you >> clarify? I would assume it is possible to >> use parts of the RDFS >> namespace without sucking in all >> dependencies, assuming we clarify >> that situation in the beginning of the >> specification. >> >> Thanks, Holger >> >> -----BEGIN PGP SIGNATURE----- >> Version: GnuPG v1 >> >> iQEcBAEBAgAGBQJU3ertAAoJECjN6+__QThfjz+8cH/3lpq+__zfMg09M01sCRIlDqi1 >> nslsOObD4ukEuioL/f9GQ1/__OZvcZVw6i09aNugsABbUHfTuFUIxsm__GA9+6r1ZM+t >> kVqzewSPhH4GFp5Gcy8x4Y0pAIEBQ6__2RRYfPNClX38eFx5e/ZJ+__xfg5HSjqzpF3r >> xVuW1+i5nge0lUJr4WF/bW/__Tj6g69TXUrXet3tNTJ1sddkxqXPo7j__BvSE1kZkBTH >> 3UsZr1yokiM6FkbxI1JJ6MIOl1BdvB__vwQaiyn38fgMjNSvTTtfvhnp3Mua8S__s4He >> 3hExQ4wUMXw0nU4ob+__71dqzvaU1o9hgRlxwgSky4gXOAmD95__U84fgpUZuVxDKWs= >> =KorL >> -----END PGP SIGNATURE----- >> >> >> >> >> -- >> Karen Coyle >> kcoyle@kcoyle.net <mailto:kcoyle@kcoyle.net> http://kcoyle.net >> m: 1-510-435-8234 <tel:1-510-435-8234> >> skype: kcoylenet/+1-510-984-3600 <tel:%2B1-510-984-3600> >> >> > > -- > Karen Coyle > kcoyle@kcoyle.net http://kcoyle.net > m: 1-510-435-8234 > skype: kcoylenet/+1-510-984-3600 >
Received on Friday, 13 February 2015 18:27:02 UTC