Re: Making progress on graphs

(for tracker, this thread has been about issue-22 for a while.)

On Fri, 2012-05-18 at 09:18 +0100, Andy Seaborne wrote:
> 
> On 18/05/12 01:17, Sandro Hawke wrote:
> > On Thu, 2012-05-17 at 18:34 -0400, Kingsley Idehen wrote:
> >> On 5/17/12 4:21 PM, Sandro Hawke wrote:
> >>> So, what should an app developer do, if they want their code to be
> >>> portable between different sparql engines?
> >>>
> >>> (answer: they should avoid using empty named graphs.  right?)
> >>>
> >>>       -- Sandro
> >> circa 2012. There are lots of developers, end-users, power-users already
> >> working across SPARQL endpoints without issue. Of all the problems there
> >> might be, empty named graphs just isn't one of them :-)
> >
> > What does that mean?   That there are much bigger problems?   Of course,
> > but does that really mean we should cause one more?
> 
> We're not causing one more.
> 
> People are reporting from experience with existing systems.
> 
> We would be causing an issue if defining RDF to be different from common 
> existing practice.

I think when we get to the point of having a test suite, and all the
people with quad-based systems realize they have to implement an
additional table of empty named graphs in order to pass the test suite,
they'll complain.     But personally, I can live with it.

On modifying N-quads to support empty named graphs, maybe this is the
time to make N-quads (or a differently-named simple, line-oriented
syntax) be a subset of trig.   Maybe each line is one of:

   <blank line>
   # comment
   s p o.           - triple in default graph
   g { s p o }.     - triple in named graph
   g { }.           - empty named graph

(This example syntax touches on another issue, which is orthogonal,
about whether the triples in the default graph need to be wrapped in
braces.   That is a separable issue, though.)

    -- Sandro

>  Andy
> 
> >
> >       -- Sandro
> >
> >
> >
> >
> 
> 

Received on Friday, 18 May 2012 15:54:14 UTC