Re: a blank node issue

On Thu, Mar 3, 2011 at 12:59 PM, Richard Cyganiak <richard@cyganiak.de>wrote:

> On 3 Mar 2011, at 07:55, Reto Bachmann-Gmuer wrote:
> > I thought the discussion was about how to best use the technologies at
> hand, not about which technologies are widely deployed.
>
> If features X has been in the spec for seven years and has had no impact on
> deployments, then it's safe to say that X is not the best way of using the
> spec.
>

Your sentence would be true, if applications would treat non-isomorphic
graphs which according to rdf semantic express the same meaning as
semantically different.  As this isn't the case I assume the spec did have
impact on deployments.


> What have you learned about RDF since 2004?
>
relevance?

>
> > I've rarely dealt with RDF applications that do not have mechanisms to
> prevent or remove redundancies.
>
> Of course, but these mechanisms have nothing to do with blank node
> leanification.
>
No? In my code these mechanism often just produce a more lean graph (by
doing specific checks and with domain specific mechanism, without doing
general purpose "leanification")


>
> > Besides creating redundant graphs starts on the first slides of an rdf
> introduction,
>
> Show me that slide set please.
>
> > a simple aggregator:
> >
> > g = new Model
> > while(true) {
> >   g load <http://.../some.rdf>
> >   sleep(...)
> > }
> >
> > If some.rdf contains bnodes g will be non-lean after two iterations.
>
> That's not a simple aggregator, it's just broken. Why would you load the
> same graph multiple times unless you expect the graph to change over time.
>

Yes, for it to be an aggregator it is supposed to aggregate different
versions of <http://.../some.rdf>, the assumption is that after sleeping for
the period ... the URI might dereference to a distinct graph. Typically this
new graph partially overlaps with the old graph.

>
> Put g into a named graph, and replace its entire contents on each load.
>
In some cases this might be the right appoach, but the strength of RDF is
that you can easily mix data from different sources. Some of these graphs
might overlap or I may aggregate the same graph twice in this case - thanks
to RDF entailment - I know I can remove some triples without changing the
semantics. Of course you can remove triples for whatever reason you like,
choose not to aggregate or not to use bnodes, my software is able to deal
with your data but if you say "this might be redundant according to this 7
year old spec, for me this is not redundant but meaningful" this would be a
hindrance to our agents understanding each other.

Reto



>
> Richard
>
>
>
> >
> > Cheers,
> > Reto
> >
>
>