Re: Linked Data Fragments: Web-scale querying

Hi Luca,

> Wouldn't it be better to *not* have the metadata return at each call
> or is it necessary in order to make hypermedia clients work?

But then only specific basic Linked Data Fragment clients
would be able to access the API.

This is exactly what I want to avoid, for two reasons:
1) I envision different kinds of clients using APIs,
    and that is only possible if those APIs are self-descriptive.
2) I envision different kinds of APIs emerging.
    What you've seen now is only the beginning of LDF;
    basic Linked Data Fragments are only one kind of fragments.
    Dozens of other possible fragments can be imagined,
    including SPARQL endpoints and subject pages existing today.

For me, those two points are the reason why the Linked Data Platform
is way too narrow. It only defines one kind of API; take it or leave it.
Clients have to be coded specifically for this API;
the API cannot evolve without a new specification.

In contrast, to _query_ basic Linked Data Fragments,
clients only have to understand the hypermedia controls.
Moreover, if they just need to _browse_ those fragments,
they don't have to understand anything, as they can just follow links.

> On that
> line, if I simply returned the data, without metadata, would that
> still be a valid LDF?

A valid Linked Data Fragment, but not a valid basic Linked Data Fragment.
According to the definition [1]:

    A basic Linked Data Fragment (basic LDF) is a Linked Data Fragment
    with a triple pattern as selector, count metadata,
    and the controls to retrieve any other basic LDF of the same dataset,
    in particular other fragments the matching elements belong to.

Note how we implemented this last constraint ("in particular…")
by adding direct links to fragments with overlapping components,
in addition to the Hydra hypermedia controls that allow to retrieve _any_ fragment.

Best,

Ruben

[1] http://linkeddatafragments.org/publications/ldow2014.pdf

Received on Friday, 21 March 2014 17:26:15 UTC