Re: Contexts (spinoff from copy and wrap rdf statements)

At 8:34 AM -0800 2002-11-26, Seth Russell wrote:
>David Menendez wrote:
>
>>At 12:08 PM -0800 2002-11-25, Seth Russell wrote:
>>
>>>Well all the triples in the graph I'm trying to come up with are 
>>>not encoded in *just one* rdf document.  That was the point of the 
>>>use case at the bottom of my last post [1].  How doe we allow 
>>>*multiple* rdf documents to assert triples to the *same*  graph ?
>>>
>>>[1] http://lists.w3.org/Archives/Public/www-rdf-interest/2002Nov/0386.html

>>Aside from having a name for an abstract graph of information about 
>>cats, what is the difference between that scenario and one where 
>>everyone just asserts facts about their own cats and agents collect 
>>the information and generate their own graphs?
>
>Naming that which you want to find is sometimes the first step in 
>finding it.  In this case the name of the graph is necessary so that 
>many agents can *cooperate* to discover it.   In fact just quoting 
>the URI of the context on your blog should help Google do most of 
>the work for you.

Why is it important that the information be asserted in a particular 
graph? To me, that just seems like an extra layer of indirection and 
uncertainty.

If Google is reading a document at catlover.org and it runs across 
the statements
	_:x rdf:type :Cat; :called "Stassi".
it can report to someone else, "According to (a document at 
catlover.org), a cat exists which is named 'Stassi'."

If the same document instead said
	{ _:x rdf:type :Cat; :called "Stassi" }
		:accordingTo <http://example.org/context/cats.rdf#ThisGraph>.
Google would have to say "According to (a document at catlover.org), 
the graph <http://example.org/context/cats.rdf#ThisGraph> asserts 
that a cat exists which is named 'Stassi'." Since no document encodes 
<http://example.org/context/cats.rdf#ThisGraph>, we can't know for 
certain whether the implication is truthful.

What does this get us? It doesn't help us find information about 
cats--in fact, it makes it *harder* to find information, because it 
increases the number of parties I have to trust in order to believe 
the information. It doesn't help us find information about a specific 
group of cats, because anyone can attribute facts to the graph.

(Stepping back a bit...)
I can see the usefulness of distinguishing between a resource and a 
particular graph that it asserts, but the naming convention you're 
proposing is problematic because it doesn't say *which* graph you're 
talking about. As an example, I have an RSS feed for my web site 
which asserts a graph, but every time I update my site, the feed 
changes and asserts a *different* graph. A naming convention won't 
work because you need to be able to say things like:

{ <http://example.org/feed.rdf> dcq:updated "2002-11-22" }
	dc:source <http://example.org/feed.rdf>; dc:date "2002-11-25".
{ <http://example.org/feed.rdf> dcq:updated "2002-11-27" }
	dc:source <http://example.org/feed.rdf>; dc:date "2002-11-27".

-- 
Dave Menendez - zednenem@psualum.com - http://www.eyrie.org/~zednenem/

Received on Wednesday, 27 November 2002 22:49:55 UTC