- From: Ruben Verborgh <ruben.verborgh@ugent.be>
- Date: Thu, 2 Jul 2015 14:38:07 +0200
- To: public-linked-data-fragments@w3.org
- Cc: Miel Vander Sande <Miel.VanderSande@UGent.be>, Joachim Van Herwegen <Joachim.VanHerwegen@ugent.be>, Laurens De Vocht <Laurens.DeVocht@UGent.be>
Dear all, The Triple Pattern Fragments interface [1] was never intended as the final answer to all of Linked Data querying on the Web. Such a final answer will likely never exist anyway, because all data publishers have to decide what is best for them. We could just go ahead and define a multitude of LD interfaces. Or just watch every party on the Web create their own interface— as is currently the case in the API landscape we see today. Instead, we aim to implement the Linked Data Fragments idea [2] by defining *features* data publishers can choose from. They indicate their choices using the Hydra Core Vocabulary [3]. For example, an API communicates to a client: “you can query by triple pattern” and/or it says "you can do full-text searches" Clients then use whichever features both they and the server support. We have been working on two such features, which we would like to spec within the Hydra Community Group. One feature is substring search [4], allowing clients to match literals. This lets us evaluate SPARQL queries with FILTERs much faster, at the expense of a slightly more complex index on the server. Another feature is approximate membership metadata [5], which lets clients evaluate queries with lots of checks faster. Both have been accepted as ISWC2015 research track papers [4,5]. If the group agrees, we would write a short spec for each of them. That way, server implementers know how to offer such an interface. Data publishers can pick any combination of features they want to support. This leads to APIs that are built from the bottom up, composed of smaller building blocks that clients can recognize [6]. Please let us know your thoughts about this. We'd love to start spec'ing this stuff ;-) Best, Ruben [1] http://www.hydra-cg.com/spec/latest/triple-pattern-fragments/ [2] http://www.hydra-cg.com/spec/latest/linked-data-fragments/ [3] http://www.hydra-cg.com/spec/latest/core/ [4] http://ruben.verborgh.org/publications/vanherwegen_iswc_2015/ [5] http://ruben.verborgh.org/publications/vandersande_iswc_2015/ [6] http://ruben.verborgh.org/publications/verborgh_kss_2015/
Received on Thursday, 2 July 2015 12:38:42 UTC