- From: Graham Klyne <GK@ninebynine.org>
- Date: Thu, 14 Nov 2002 16:14:13 +0000
- To: "Richard H. McCullough" <rhm@cdepot.net>
- Cc: "Danny Ayers" <danny666@virgilio.it>, "RDF-Interest" <www-rdf-interest@w3.org>
At 06:06 AM 11/14/02 -0800, Richard H. McCullough wrote: >I'm having trouble understanding your notion of "context". As I am yours. The term has been much abused (and I don't excuse myself). Because it's working software, Notation 3 is probably a useful common ground, with its notions of formulae (roughly: collections of statements) and context "Context is the relationship between a statement and the formula it is directly part of" [1]. I don't fully understand the last bit, but the idea of context as a relationship rather than a thing seems to sidestep the discussion of forward/backward view of these things. #g -- [1] http://www.w3.org/DesignIssues/Notation3 >Paraphrasing your three properties: > Resource has context = Statement > Statement has contains = Resource > Bag has contents = Resource > >This seems backwards to me. I think of context >as follows: > Statement has context = cname > cname is List of Statement > >Remarks: >1. Although a Bag of Statements will work in some >cases, I think that a List is necessary in general. >2. If you want to include actions (as opposed to static >Properties), then context should include space and >time. >3. In my KR language, I put the context before the >Statement (I call the static context "view") > at view = cname { Statement } >My static Statement is written as > subject has predicate = object >============ >Dick McCullough ><http://www.volcano.net/~rhm>knowledge := man do identify od existent done >knowledge haspart list of proposition >----- Original Message ----- >From: <mailto:GK@ninebynine.org>Graham Klyne >To: <mailto:danny666@virgilio.it>Danny Ayers >Cc: <mailto:www-rdf-interest@w3.org>RDF-Interest >Sent: Wednesday, November 13, 2002 4:59 AM >Subject: Re: Contexts (not again!) > > > >I think the "dark triples" approach fizzled out. My take is that we're not >ready to standardize context mechanisms yet, but still have hopes of >prototyping my ideas in this area, which aren't vastly different from what >I think you're describing. I think that reification, or a variation of it, >can be used (in a prototype implementation) to encode the triples that >aren't asserted. > >In the longer run, a standard solution may call for something more >"hard-wired", with scope for optimization. I think this might come about >without invalidating/isolating the >prototype approaches. > >#g >-- > >At 10:59 PM 11/2/02 +0100, Danny Ayers wrote: > > >Hi folks, > > > >Did any kind of consensus, or even decision (!?) result from Pat's 'dark > >triples' suggestion [1] etc. earlier in the year (or any other of the > >familiar context discussions)? I've had a look through the archives and as > >usual the threads are hard to follow. I'm wondering because I'm running up > >against this thing again. > > > >If there isn't anything sorted or on the cards in this area, I'd appreciate > >comments on the following first crack hackiness for a context vocabulary. > >I've not really got a grip on the reification angle with it yet, but the > use > >I'm after is really just to be able to tag triples (make 'em quads in > >memory), and it'd be nice to do it in a moderately sound fashion. > > > >Just three terms (the pseudo-schemas are undoubtedly way out) : context, > >contains, contents > > > >*context* - a group of statements (identified collectively by a single URI) > >with which a particular statement can be associated. In practice this would > >usually be > > > >[triple]-context->[RDF file] > > > >Property "context" > > domain Resource > > range Statement > > inverseOf contains > > > > > >*contains* - the other way around, > > > >[RDF file]-contains->[triple] > > > >Property "contains" > > domain Statement > > range Resource > > > > > >*contents* - a list/collection whatever of (references to) the > statements to > >be identified by a given URI (i.e. the triples in a file) > > > >Property "contents" > > domain Bag > > range Resource > > > >[RDF file]-contents->[s1, s2...] > > > >The first of these is probably all that I'd need, but the second > insisted on > >coming along. The third heard there was a party. > >When I started thinking of a way around this, the first thing that came to > >mind was a Context class, akin to a collection/bag, instances of which > could > >be used to identify a file but with this it seemed to get messy a lot > >quicker... > >I'm pretty sure I'm badly conflating the unreified/reified triples here, > and > >it does seem like it goes a bit beyond what can be expressed in RDF(S) > alone > >(i.e. a minilayer on top) but I'm hoping that something usable won't be far > >away. I'm willing to bet there's something along these lines already, but I > >can think of worse ways to spend a Sunday evening. > > > >Cheers, > >Danny. > > > >[1] > <http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Mar/0253.html>http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Mar/0253.html > > > > > > >----------- > >Danny Ayers > > > >Semantic Web Log : > >http://www.citnames.com/blog > >------------------- >Graham Klyne ><<mailto:GK@NineByNine.org>GK@NineByNine.org> ------------------- Graham Klyne <GK@NineByNine.org>
Received on Thursday, 14 November 2002 11:22:37 UTC