Re: life without dataset semantics

On 09/19/2012 10:32 AM, David Wood wrote:
> On Sep 19, 2012, at 10:06, "Peter F. Patel-Schneider" <pfpschneider@gmail.com> wrote:
>> On 09/19/2012 10:02 AM, Sandro Hawke wrote:
>>> On 09/19/2012 09:48 AM, Peter F. Patel-Schneider wrote:
>>>> I'm not convinced that there is any need to restrict properties like sendCorrectionsTo to datasets.
>>>>
>>> How else could you define/document it?     In writing the property documentation, I find myself needing some way to talk about those intended triples (the ones containing the information that the corrections are about).  Without a dataset or some global relation underlying dataset semantics, I don't know how to do that.
>>>
>>>     -- Sandro
>> You could just say that it is a relation between the name/location of a graph and something else.  in a dataset it would be the named graph in that dataset (or not).  Elsewhere it would refer to the graph at a location.
> Sandro, are you looking for a URI that you can use as the rdfs:range for the property sendCorrectionsTo?

That might be useful, but it's not my focus.  I'm just thinking about a 
one-paragraph or one-page explanation of the sendCorrectionsTo 
predicate.   How could you explain to all the people who might use this 
predicate what exactly it means.   These are folks who are releasing 
data feeds from government agencies, research labs, etc, etc, and the 
folks writing software which uses those feeds.   They need to all have 
the same picture in their heads, more of less, of wh  (This was what I 
proposed in [1].)at sendCorrectionsTo means, in both the common cases 
and the corner cases.

I think if the RDF WG lays the right groundwork -- and we've very very 
close -- then other people can easily define,  document, and begin using 
sendCorrectionsTo and a few dozen other very interesting and very useful 
predicates.  (We might also take a stab at some of them in a WG Note.)   
If we don't lay the right groundwork, I'm not sure what will happen.

The right groundwork certainly includes a simple definition of datasets 
and some terminology around them (maybe the stuff from SPARQL is 
fine).   It should also help people make sense of situations like the 
predicate being used outside of a dataset, or in a named graph in a 
dataset vs the default graph.     Frankly, I can't quite yet make sense 
of those corner cases, with this zero-semantics approach.  They do make 
sense to me, however, if there's some kind of global relation between 
things denoted by graph names and the graphs, so that's the sort of 
thing I keep leaning towards.

     -- Sandro



> Regards,
> Dave
>
>
>> peter
>>
>>
>
>

Received on Wednesday, 19 September 2012 14:58:20 UTC