Re: Proposal: The "POI as a simple collection of links" data model

OK...Thomas seems to be ok with this.

And I had a quick irc chat with Dan about this before he left for
holiday and he's ok with it too.

So if there's no objections, I'll work this up into a more detailed
example.


roBman


On Fri, 2011-08-05 at 02:58 +1000, Rob Manson wrote:
> Hey Dan,
> 
> I'm willing to have a crack at mapping this out.  
> But any input you can provide would be really helpful 8)
> 
> Here's an abstract attempt at a very high level.  
> Are these the sort of relationships you meant?
> 
>         <pois> <contain> <poi>
>         <pois> <crs> <wgs84-3d>
>         <pois> <metadata key> <default content>
>         <pois> <default script> <src uri>
>         <pois> <default style> <src uri>
>         
>         <poi> <___location> <geo uri> or <poi> <___location> <uri> or <poi> <___location> <gml point>
>         <poi> <alternate> <uri> <- what's the best way to model attributes?
>         <poi> <metadata keywords> <content>
>         <poi> <dc:title> <value>
>         <poi> <dc:publisher> <value>
>         <poi> <script> <src uri>
>         <poi> <style> <src uri>
>         <poi> <change to state> <date/time>
>         <poi> <gr:has opening hours specification> <gr:opening hours specification>
> 
> 
> So as you hinted at:
> 
>         <poi> <<link type>> <<value>> <- set a value
>         <poi> <<link type>> <<href>>  <- relate to another http accessible thing
> 
> 
> Obviously these need to use valid URIs...and to use correct notation.
> But I just wanted to check I was modelling the right aspects first.
> 
> Does this type of model guarantee a <poi> or <pois> could then be
> injected into any rdf/xml entity representation (or any other
> notation/serialisation) just like foaf and dublic core metadata, etc.
> can too?  And likewise any foaf or dublin core metadata can be packed
> into <pois> and <poi> too?
> 
> Being able to store them all in existing triple stores sounds like a
> great advantage too...
> 
> NOTE: See my question above
> 
>         "what's the best way to model attributes?"
> 
> Either at the predicate or object level.  Or are you just meant to
> define a triple for each where the predicate is the attribute?
> 
> 
> 
> roBman
> 
> 
> 
> On Thu, 2011-08-04 at 16:01 +0200, Dan Brickley wrote:
> > On 4 August 2011 15:56, Raj Singh <rsingh@opengeospatial.org> wrote:
> > > I love the "POI as a simple collection of links" data model in theory. It's very clean conceptually, and that's the way I've been thinking of the larger POI "picture" also. However, I don't think it's the best way to implement.
> > >
> > > I think we still need a few primitives. The only things that are in the model now that don't fit your links model are the label, some time fields, and the ___location. The label and the time fields are so natural to just stick in there. That brings us to ___location. The data URI seems awkward to me and likely to turn people off of the spec. However, I'm open to other opinions.
> > 
> > I like it too. Seems pretty much isomorphic to RDF, btw. Can it be
> > written out explicitly as a set of triples, eg. an unordered
> > collection of factoids that take one of these forms:
> > 
> > <thing1-URI> <propertytype-URI> "value " .
> > 
> > and
> > 
> > 
> > <thing1-URI> <propertytype-URI> <thing2-URI>.
> > 
> > ...?
> > 
> > If so, it should be possible to write out this datamodel using any RDF
> > notation, store/query it in any RDF database, etc.
> > 
> > cheers,
> > 
> > Dan
> > 
> >

Received on Thursday, 18 August 2011 11:54:25 UTC