W3C home > Mailing lists > Public > semantic-web@w3.org > March 2014

Re: Linked Data Fragments: Web-scale querying

From: Ruben Verborgh <ruben.verborgh@ugent.be>
Date: Wed, 19 Mar 2014 15:14:57 +0000
Cc: Linked Data community <public-lod@w3.org>, "semantic-web@w3.org Web" <semantic-web@w3.org>, Pieter Colpaert <pieter.colpaert@ugent.be>
Message-Id: <CA78690B-4880-46D6-B15D-D5BCD6A68DB4@ugent.be>
To: Luca Matteis <lmatteis@gmail.com>
Hi Luca,

> Just finished reading the paper. Really great stuff. The idea of
> splitting a single resourceful SPARQL request into multiple
> fine-grained requests was always attractive to me and I've also
> thought of something similar (http://lmatteis.github.io/restpark/). I
> applaud you and your team for coming up with a formal client algorithm
> as well as a server implementation for offering this functionality.

Thanks and some of the ideas seem similar to Restpark,
especially the URI template and the examples,
which come down to concrete applications of our algorithm.

The main difference in the design process, I think,
is that I started looking at it from the client perspective:
what affordance does a client need to solve queries?
Incorporating the total triple count and links to other fragments
are very important to allow the right client-side decisions.

> I wonder, given a Linked Data set, if you could easily and generically
> wrap it with a basic LDF interface. Sort of how Pubby wraps SPARQL to
> expose Linked Data, maybe there can be a wrapper for a Linked Data
> site to expose basic LDF.

In addition to Pieter's answer, who correctly points at the implementation
of third-party tools, I would also point to our open-source basic LDF server:

As you can see, it supports many back-ends out-of-the-box.
This includes SPARQL endpoints, but also other triple sources
such as Turtle files, and even other basic LDF sources.

An interesting back-end is HDT (http://www.rdfhdt.org/),
which we use to offer http://data.linkeddatafragments.org/dbpedia.

> * I build a nice site and my data is available as HTML, marked up with RDFa
> * Database is not a triple store, just regular MySQL

So the easiest way to publish this as fragments could be to extend Datasource.
For an example on how to do this, see

> Where I see LDF being a *huge* deal is that I could use something to
> wrap my RDFa pages and expose a basic LDF server, without having to
> change any of my technology stack for my app. This could potentially
> allow thousands of RDFa providers to expose querying functionality
> with minimum effort.

Indeed, and that's exactly what we aim for with basic LDFs:
low-cost, queryable publishing of Linked Data.
Let us know how it works out, or if we can help!


Received on Wednesday, 19 March 2014 15:15:35 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:45:36 UTC