Software InfrastructureThough M-Zones will be conducting research into adaptive and policy-based techniques for performing
adaptive smart space management, this can only achieve results if supported by a clear view of
common software services and capabilities that will be available for application developers. We call
this set of common software services, technologies and capabilities the software infrastructure. In any
particular domain of computing, the software infrastructure is constantly developing, with new
functionality being added, often at higher levels of abstraction than before. A good indication that some
piece of technology is accepted widely enough to become part of the ‘infrastructure’ is when associated
interoperability standard emerges. Sometimes the emergence of such standard is the catalyst for
widespread uptake and ‘infrastructuralisation’ of a technology. This section reviews aspects of the state
of the art in the software infrastructure that appears to be in place for ubiquitous computing. M-Zones
moves forward with the assumption that such an infrastructure will form the basis of our research into
adaptive smart space management, but also with the expectation that technologies that are at the
research and development stage now will become part of the software infrastructure as the project
progresses. Equally, improvement of the infrastructure will be an active research area in M-Zones.
This section addresses software infrastructure through three state of the art whitepapers:
Context information is important both to enable ubiquitous computing systems to adapt to a changing
physical and computing environment and to support them in understanding changing user needs and the
best corresponding course of action to take is any particular situation. Context information includes
computing context, the user context and the physical context as well as historical information about
these. As any form of adaptive smart space management system will need access to context
information, inclusion of context management capabilities into the infrastructure is a sound
architectural decision. Context management is already in place in a rudimentary sense, for example in
the way desktop office application use information on past and current activities to provide context
sensitive help or in the way web browsers and server collaborate to detect if someone has visited a web
page before. Much work to date in relation to ubiquitous computing has focussed on location detection
as a primary source of context information, however the Context Management white paper points out
that a much broader view of context information is required if context management is to be offered as
an infrastructural service. This will require standardisation of context information through adoption of
common models and service for manipulating context information, rather than the wide range of highly
application dependent models and service currently available. This will require some commonality in
how context data is collected from sensors, applications and other parts of the infrastructure, but also
strong privacy control mechanisms to provide people with the confidence to allow personal context
information to be shared with encountered applications and ubicomp environments.
A core assumption that is essential to developing adaptive system that can dynamically integrate
components from multiple developers and operated in multiple organisational domains, is that
components only interact through well defined interfaces. This is known as a service-oriented
approach, where such well defined interfaces, or services, are reflective, i.e. their capabilities can be
examined at run-time. The level of expressiveness and semantics used in a reflective service
description determines the level of automated decision making that can be performed by an adaptive
system. The assembly of services into a more highly functional aggregate service is called service
composition, and a principle aim of adaptive smart space management of to perform service
composition is as dynamic and seamless a manner as possible. The Service Composition white paper
discusses the various activities that are involved in service composition, including the description of
services at various levels of abstraction, the advertisement of services and the use of brokers to locate
services that best fit requirements and trust requirements for using foreign services. The automated
composition of services is still a very open research area and centres on the combination of various
reasoning techniques with common mechanisms for expressing semantics via ontologies.
Standardisation of ontology language is underway in ISO (Topic Maps) and W2C (based on
DAML+OIL). Building on these DAML-S utilises concepts from workflow and web service flow
languages to provide a semantic service description and composition language.
Underlying application development in any distributed computing environment is the concept of
middleware, which aims to provide the application programmer with a portable API and a set of
common services. Middleware systems and standards are very mature, with CORBA, Sun’s Java RMI
and Microsoft’s .NET being the most widely used. A wide range of general purpose and application
specific services are available for these platforms and interoperation solutions between platforms are
widely available. In addition, most of these platforms now take a component-oriented approach to
application to ensure ease of portability and deployment on different platforms. Middleware system
also typically support exposing interfaces as web services so that these services can be easily accessible
over the Internet (most middleware platforms cannot negotiate unmodified firewalls). Despite this
maturity the middleware area is developing fast along a number of axes described in the Middleware
whitepaper, which may be useful for adaptive smart space management. One specific example is Jini,
which is Java based middleware aimed at ad hoc communication between a wide range of personal,
office and domestic devices. Another broader movement is that of intelligent mobile agents. These
primarily take advantage of the code mobility offered by Java to allow active software components,
called agents, to move around a network in order to accomplish some task. As agents will have no a
priori knowledge of other agents they will encounter and collaborate with in accomplishing tasks, there
is quite an amount of research on agent communication mechanisms where capabilities are exchanged
and mutual activities negotiated. This makes agent very applicable to the highly dynamic computing
environments of smart spaces. Finally there are some standardised application specific APIs that may
be relevant to the integration of smart spaces with other system. Parlay is a good example, being an
industry API intended to make management and control functionality of telephony systems available to
other systems.
|
