[foaf-dev] is foaf graph limited to obtaining its bearer document from the uri of the named object in the graph?
pwilliams at rapattoni.com
Tue Oct 27 11:18:50 CET 2009
Henry taught me that its cute that a uri naming object X in a foaf graph of triples can be resolved to the foaf document stream at a real web server (given HTTP rules of getting concrete resources such as document streams). This I have internalized, and it now represents to me *the* model of identification/naming/address/locating that is part of the semweb metamodel. IN short, http scheme URI names can name arcs and nodes in a graph, and an HTTP GET on that name can deference the file that represents the graph in the world of bits and bytes.
But, is the semweb limited to that model of naming/addressing/locating? I realize its almost heresy to use other than the URI identifier model leveraging http scheme URIs for naming. But is semweb formally so limited (in the underlying rdf semantics, for example)?
I built myself a simple graph from a type system conformed only of foaf classes, foaf properties (and rdf entities). It described a host (a foaf agent), with a web server (another foaf agent), hosting a file (a resource), which is a foaf document, in which a graph of objects describe other foaf agents and foaf groups. The graph describes a locator service for de-referencing the uris of other foaf files, pertaining to the members of certain groups specified to associate with the foaf agent known as my host PC (which has a structured name assigned to it known as a (DNS-listed) domain name).
Now, if I were to go around telling folks that the foaf file of some friend is NOT to be obtained by simply following a webid name, but is to be located by doing an HTTP GET on the URI http://domainname/getfoaffiles?uri=<webid<http://domainname/getfoaffiles?uri=%3cwebid>>, would that locator scheme be "contrary to semweb" doctrine? Is it contrary to foaf doctrine or linked data doctrine?
Or, is using a locator service just as valid a way of telling folks HOW to deference a webid - as is simply getting the document implied by the webid name using HTTP GET directly?
When I read the book on semantic web for the working ontologist, I remember it discussed the notion of naming aggregators, semweb services that could construct non URI names for objects (based on notions such as assigning a structured name based on category maps. These non URI name forms MIGHT be merged with URI stems to then make a URI - of the form http://namer/getter.aspx/a/b/c (where a/b/c is the structure name assigned by the naming aggregator according to its categorization). Given this example, it seems reasonable that my foaf file MIGHT locate a friend's foaf file not using a simple http uri (or a webid), but by defining its name/address using a particular referral/locator service (itself described my foaf file).
If I am a foaf agent that in reality is a simple internet host, on a domain, my foaf file might be defining to th world how users associated with my domain are not only declared but how one can get their foaf file (by using some query-capable naming aggregator service at URI X).
If this doesn't really make any sense, just ignore. Its an exercise for me - tyring to "think" semweb - simply to ensure I understand its metamodel and any limits.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the foaf-dev