A Simple Architecture for Delivering Context Information to
Mobile Users
Sigrid Steinholt Bygd s, P l S. Malm and Tore Urnes
Telenor Research and Development
P.B. 83, N-2027 Kjeller, Norway
+47 63 84 84 00
sigrid-steinholt.bygdas@telenor.com
Mobile devices running context-aware applications are in need of data sources delivering
context information. We envision that context information data sources will be offered by
third-party providers, and will likely be distributed across the Internet. Context
information providers will prefer to use small and simple databases that do not require
complex data modeling, configuration, and administration.
Context-aware applications require that database systems be able to run application-
specific queries, often involving tailored operators. Unfortunately, it is impractical to
manually deploy the tailored operators of each application on the necessary remote data
sources. A solution to this problem is to have applications dynamically move query
operator code to the remote database sites, as described in [1].
In this paper we briefly describe an architecture for delivering context information to
wireless mobile devices from data sources on the Internet. The architecture is based on
code mobility and the conceptually simple tuple space technology.
Context-aware information retrieval
In [2] Brown proposes a model for context-aware information retrieval. In this model,
each piece of context information, i.e. a piece of information augmented with context
data, is represented by a set of fields, each field being a name/value pair. Based on this
model, a rule for matching the present context with pieces of context information can be
defined. A matching rule defines the fields that are to be compared with the present
context and the conditions for a field value match. For example, check all fields with
name temperature and location, and return a temperature match if the value of the
temperature field is the same as the present temperature, and a location match if the
present location is closer than 100 m from the value of the location field. Matching rules
may change dynamically, for example if a user removes a temperature sensor the
application should no longer take temperature into account when retrieving information.
Tuple space technology
Tuple space technology [3] provides a global communication buffer for distributed
applications, ensuring loose coupling between information sender and receiver. In a tuple
space, information is organized in tuples. A tuple is a vector of typed values called fields.
Storage of tuples in a space is persistent and the operations performed on the tuples are
transaction secure. The search for information in a tuple space is made associative by
using templates. A template is similar to a tuple except that some of the fields may not
have values. A tuple matches a template if they have the same number of fields and each

corresponding field has the same type and value (or just type for template fields without
values). A template that looks like this: [27.09.2000, string, conference ] matches the
following tuple [27.09.2000, Bristol , conference ].
In our view, tuple space technology offers a natural (and simple) way of implementing
Brown s information model. This is due to the fact that a set of name/value fields can be
represented by a vector of type/value fields (a tuple) where the field names are values of
type string, like [ temperature , 17.5]. Tuple space technology does not, however,
provide sufficient flexibility to realize the matching rule semantics. What is needed is an
extended tuple space implementation that facilitates dynamic customization of the tuple
space matching algorithm. Recently, researchers at IBM provided such an extended tuple
space system called T Spaces [4]. Based on Java and its support for code mobility,
T Spaces offers dynamic uploading of application-specific query operators (matching
rules). T Spaces also offers traditional database features while retaining the simple
conceptual model of tuple spaces.
Proposed architecture
We propose an architecture for distributed context aware applications based on code
mobility and tuple space technologies. Under this architecture, context information data
sources are implemented using T Spaces and made available on the Internet. Context-
aware applications running on wireless mobile devices dynamically install their query
operators at the appropriate data sources (for now, discovery of available information
sources is realized in a heuristic fashion). For low bandwidth situations we recommend
deploying triggering engines and filters on proxy servers at the edge of the wired
network, hence minimizing traffic over the wireless link.
Proxy
Trigger
engine
Tuple
space
Tuple
space
:
Figure: A simplified illustration of the architecture.