W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > July to September 2010

Re: Mutability and graphs [was: Re: page about the term "named graphs"]

From: Axel Polleres <axel.polleres@deri.org>
Date: Fri, 23 Jul 2010 22:33:15 +0100
Cc: "SPARQL Working Group" <public-rdf-dawg@w3.org>
Message-Id: <2902884B-3C06-44A5-867C-151B24D8BF9E@deri.org>
To: "Andy Seaborne" <andy.seaborne@epimorphics.com>

On 23 Jul 2010, at 17:56, Andy Seaborne wrote:

> 
>  > ----------------------------------------------------------------------
>  > Dataset(new): Given RDF graphs (referred to by symbols)
>  > <G_0>, .... ,<G_l>,<G_n1>, ....<G_nm>  a dataset DS is the following
> set of
>  > (mathemetical) sets,
>  >
>  > DS = {<G>,<G_n1>, .... ,<G_nm>}
>  >
>  > where G = {<G_0>  merge .... merge<G_l>  }
>  >
>  > and merge is RDF merge in the sense of rdf-mt.
>  > We also denote the symbol referring to a graph as "name" and particularly
>  > call the graph<G>  "default graph".
>  > ----------------------------------------------------------------------
> 
> Why do you need to redefine RDF dataset?  

See my previous mail... I had the impression that several people were unhappy with named 
graphs as we have them (pairs graph-name). My idea/suggestion was to try out
another definition which might be more what these people expect in an 
upwards compatible way with SPARQL1.0, i.e. ...


> You don't make use of this
> definition anywhere.  What visible difference are you getting at?

... I wanted to attempt a redefinition with exactly that "feature" of no visible difference.

> If
> there is no visible difference, why are you proposing a change? You
> don't use the definition for your graph store proposal where you just
> use "dataset".

It seems we may have a chance to reach convergence in this discussion 
then on the graph store definition at least, which is something!
(let me still comment on the alternative dataset definition comments in the following, 
as I still believe it works, without firmly arguing at this point for adopting it)

> The definition does not make sense to me. What is <G_0>? Is it an IRI?

<G_0>, .... ,<G_l> are meant to be symbols referring to Graphs and these symbols
are IRIs . no, in your terminology you used S_0, ... S_l, the only difference is that 
I proposed to use IRIs to denote the symbols for the graphs directly.

> Or a name for the purposed of this definition?  If it's an IRI, then
> 
>  > DS = {<G>,<G_n1>, .... ,<G_nm>}
> 
> is a set of IRIs, right?.

well, then { S0, .... Sn } would also be a set of symbols, but no, I meant 
the set of *graphs* referenced by these symbols.

>  How does the spec get a graph in order to
> query it?  In SPARQL 1.0 it's Map IRI->graph.  (A Map is a Set of pairs
> with a look function from "key" the first in a pair to "value" the
> second in a pair.  In Java you can even iterate over a Map as a set of
> pairs.)

In Java/programmatically, that would stay exactly the same, as said above, the definition 
was solely meant as attempt to move from the current named graph definition to something 
more acceptable for those inconvenient with that definition - with exactly the purpose that 
it causes no visible difference.

> 
> 
> How can RDF Graphs, which are sets, have names.

symbols, not names

> Sets don't have "names"
> other than in the self-denoting sense that literals do.

I can define the set S = {t1, .... tn}, so why shouldn't I be able to 
define the set <g> = {t1, .... tn} likewise - understanding IRIs as symbols.

> 
> Suppose:
> GET http://faraway/resource -> "1"^^xsd:integer
> 
> today
> and
> 
> GET http://faraway/resource -> "2"^^xsd:integer
> 
> tomorrow.
> 
> so http://faraway/resource is not a permanent name for the number 1.

because that http://faraway/resource is rather a graph store consisting of a single graph 
than a graph? (cf. previous mail)

> 
>  > DS = {<G>,<G_n1>, .... ,<G_nm>}
> 
> is that a set of symbols or a set of graphs?

the latter, where the graphs are dereferenced by the symbols, see above.

> Or a set of symbols and graphs?
> 
> How does this work for GRAPH, FROM and FROM NAMED?
> 
>  > G = {<G_0>  merge .... merge<G_l>  }
> 
> This breaks SPARQL 1.0.
> 
> The default graph IS NOT the RDF merge of the named graphs.  It is a
> graph without constraints. /Some/ systems may make it the RDF merge of
> the named graphs

you misunderstood me, <G_0> ... <G_l> are not named graphs, but those graphs
that make up the default graph (by the FROM clauses)
whereas <G_n1>, .... ,<G_nm> are the named graphs... maybe that's our misunderstanding.


Hope that clarifies matters,
Axel

> (usually, they don't, they make it the union but that's
> OK with some wiggle room about a priori knowledge of the domains of
> discourse - /pace/ Steve).
> 
> BUT this is NOT required by the spec for SPARQL 1.0.
> 





>         Andy
> 
> 
Received on Friday, 23 July 2010 21:33:45 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:43 GMT