- From: Harry Halpin <hhalpin@ibiblio.org>
- Date: Tue, 28 Mar 2006 12:19:39 -1000
- Cc: Richard Cyganiak <richard@cyganiak.de>, semantic-web@w3.org
I'm going to note that in the parallel universe of "vernacular" XML development, this seems to be similar to the debate about versioning and namespaces. To summarize, the namespace by itself does not define the application semantics of an XML vocabulary, and XML specifications certainly don't for most of the non-W3C approved XML documents. So it requires "out-of-band" information sometimes to determine what someone *means* by the use of an XML element/attribute and their provenance information. One can (but does not have to) retrieve a namespace document to get authoritative information about an XML vocabulary as given by the namespace URI it uses. One could do something similar with applications of RDF/OWL, and put default versioning and provenance information in the namespace document (which could contain links to a FOAF file, RDF(S), and so on) of the namespace URI used by a particular vocabulary. Most of these namespaces seem just to have RDF schemas in there, but you could be put more there, like provenance. It seems straightforward, and there's even a RDF mapping for RDDL by DanC, but I'm not sure about what RDF provenance vocabularies are out there... -harry Ora Lassila wrote: > On 2006-03-28 13:24, "Richard Cyganiak" <richard@cyganiak.de> wrote: > > >> On 28 Mar 2006, at 19:31, Ora Lassila wrote: >> ... >> >>>> Is it correct that out-of-band information (e.g. a web page stating >>>> "All these files are up-to-date", or some nonstandard extension of >>>> RDF) is necessary before an agent can safely act upon any RDF >>>> statement? >>>> >> ... >> >>> IMHO, this is a question that could be asked about *any* document >>> that has >>> been published, not just RDF documents. The question is more about >>> *who* is >>> asserting. I could assert that, say, the Moon is made of cheese. >>> Whether >>> someone else chooses to *believe* this is another matter. Whether I >>> assert >>> that in RDF or in natural language is not so relevant. >>> >> Right. When you assert this in natural language, I can use out-of- >> band information ("common sense") to decide wether to trust your >> statement or not. >> > > Well, I guess common sense is only *one* of many ways how to make those > decisions (e.g., I am not sure that common sense would have been enough to > evaluate DanBri's example about the weapons of mass destruction). > > >>> The key responsibility (again, IMHO) of "Semantic Web agents" is to >>> make >>> decisions (and inferences) about what information to trust, to use, to >>> discard, to keep but not trust, etc. >>> >> That makes a lot of sense. Am I correct when I say that RDF and OWL, >> at the current state of standardization and common practice, don't >> provide a solution for this trust problem, and application developers >> are on their own? >> > > Yes. I don't think RDF and OWL should, per se, even provide a solution, > given that the required mechanisms can be application-specific. Could one > potentially *model* some of those mechanisms using RDF and/or OWL? Why not. > > - Ora > > -- -harry Harry Halpin, University of Edinburgh http://www.ibiblio.org/hhalpin 6B522426
Received on Tuesday, 28 March 2006 22:19:49 UTC