Re: [Gzz] Wrapper mini-API to RDF

[Benja Fallenstein]

Alatalo Toni wrote:
> On Tue, 18 Mar 2003, Benja Fallenstein wrote:
> > I think that would be exactly backwards: First make your experience with
> > the APIs out there, then decide whether you need your own... :-)
>
> i read this so that what follows must not necessarily be made

Yup.

> > Fast access style: for views, using a special kind of iterator to avoid
>
> isn't there the thing about using impartial/incomplete data?

I don't understand this, but from what follows, I guess you were 
off-track here already :-)

> (also due to
> delays in a networked implementation?)

This is not for a networked implementation, this is for an in-memory 
impl: The thing views and bindings are implemented on top on, something 
you interact with at interactive speeds.

We need the 'fast access style' because creating a lot of objects slows 
our views down-- because we need to traverse a lot of connections 
(hundreds to thousands) in 30 milliseconds. You're not talking networked 
implementations in these time frames :-)

> >         iter.lookup(subj, prop, obj);
>
> .. as the lookup time varies?

Nope. Absolutely nothing to do with that.

iter.lookup(subj, prop, obj) is a different interface to the same data 
that graph.get(subj, prop, obj) returns -- a set of triples. With 
iter.lookup, you can use the same iterator object to retrieve more than 
one set of triples, like,

    iter = graph.getIterator();

    iter.lookup(something);

    // process first set of triples
    // iterating through the triples using iter.go()

    iter.lookup(something else);

    // process second set of triples
    // iterating through the triples using iter.go()

Ok?

> or is this stuff handled on a different level?
> i haven't read the dtn related stuff mentioned.

Ohmigod. This is for interactive speeds, where lookup times are counted 
in nanoseconds or so. DTN is for networks where TCP cannot be used 
because a roundtrip may take hours, days, weeks! :-)

-b