Sesame RDF Repository Extensions
for Remote Querying
Simon Schenk1 and Josef Petrák2
1Department of Computer Science,
University of Koblenz–Landau
sschenk@uni-koblenz.de
2Department of Information and Knowledge Engineering,
University of Economics Prague
petrakj@vse.cz
Sesame RDF Repository Extensions for Remote
Querying
Simon Schenk1 and Josef Petra´k2
1 Department of Computer Science,
University of Koblenz–Landau
sschenk@uni-koblenz.de
2 Department of Information and Knowledge Engineering,
University of Economics Prague
petrakj@vse.cz
Abstract. The Semantic Web is envisioned to be a networked environ-
ment of machine understandable and interliked information. Despite this
vision, however, most Semantic Web applications working on datasets
from multiple sources today replicate all neccessary to a single repository.
This repository is then used for query answering. We present mechanisms
to formulate views on RDF datasets and to evaluate queries against dis-
tributed datasets, which do not require data replication, but work in a
truly distributed fashion.
Keywords: RDF, SPARQL, Distributed Querying, Views, Networked Graphs
1 Introduction
Data reuse and integration is the basic idea behind the Semantic Web effort.
However, the extremely large and highly distributed setting of the Web makes
reuse and integration a difficult task. Although the basic technologies of the
semantic web (URIs to denote things, shared ontologies) are well suited for a
distributed environment, data reuse and integration today usually takes place
in centralized settings. Data integration often is done by replicating all data to
a single repository and connecting data sources using procedural code.
Among others, replication leads to problems with (1) staleness, as data is
frequently updated, (2) scaleability, as (combinations of multiple) large datasets
can not easily be handled and (3) access rights, as not all data will be available
for copying.
The Semantic Web consists of of machine understandable information, op-
posed to the machine readable, but not machine understandable Web we have
today. RDF [2] is the basic data model of the Semantic Web. In RDF informa-
tion is represented as statements of the form (subject, predicate, object).
These statements form a graph, when subjects or objects are used in multiple
statements. As most other resources on the Web, all components of a state-
ment are uniquely identified using a URI. On top of this basic model, various
inferencing mechanisms, e.g. RDF-Schema [2] reused.
c© Václav Snášel (Ed.): Znalosti 2008, pp. 375–378, ISBN 978-80-227-2827-0.
FIIT STU Bratislava, Ústav informatiky a softvérového inžinierstva, 2008.

376 Simon Schenk and Josef Petrák
Named graphs are tuples of a name (again a URI), and an RDF graph.
They allow to refer to a piece of information (the graph) and, using the name
as subject or object, to express information about the information contained in
the graph. The most commonly used query language for RDF is SPARQL [1].
SPARQL is based on graph pattern matching. A pattern basically is a graph
template formulated by using variables in subject, predicate or object positions.
The values obtained during graph pattern matching can be used to again create
valid RDF — which may but needs not be different from the input graph. Hence,
SPARQL is a powerful mechanism for information extraction and reuse.
2 Proposed extensions
We extend the RDF model and SPARQL query evaluation in two ways:
First, we define Networked Graphs [3]. A Networked Graph basically is a
named graph, which may be defined as a view on one or multiple other RDF
graphs. This allows for the easy reuse of existing data, without a need to replicate
it. Networked graphs can be formulated using an RDF and SPARQL based
syntax. They can easily be exchanged and are upwards compatible with existing
RDF infrastructure.
Second, we extend the SPARQL query model, such that the operators of
SPARQL take the repository into account, where the data they operate on is
stored. If this is different from the local repository, parts of a query are forwarded
to be evaluated at other repositories, so called SPARQL endpoints. The com-
munication takes place using the standardised SPARQL protocol. Hence, data
at remote repositories can transparently be accessed without the need for data
replication.
2.1 Sesame architecture
We implement our extensions based on the Sesame RDF repository framework3.
The Sesame framework architecture consists of the following components: The
repository is the high–level object which defines the application interface and
offers various methods to query data, upload files, extract or manipulate the data.
It abstracts from the internal implementation, which consists of so called SAIL
objects. So called SAIL (Storage And Inference Layer) components abstract
the repository implementation from details of storying, or inferencing. A SAIL
implements either actual storage of RDF data, of inferencing capabilities based
on some underlying storage. Using a set of SAILs for storage and inferencing, the
actual behaviour of a Sesame repository is defined by a stack of SAILs. Sesame is
shipped with various SAILs for in-memory or on disk storage and for inferencing
using RDFs and custom rules. The extensions described here are implemented as
two SAILs, which add additional querying and inferencing capabilities. Hence,
they can be combined with features implemented in other SAILs.
3 http://www.openrdf.org/