Software Retrieval Service - SRS
Last update: March 12, 2002
Explanation of service
It is an alternative procedure for users to retrieve and install software. The
Software Retrieval Service, based on the use of an ontology and the agent
technology, allows users to retrieve and install software in an easy and
efficient way. Easy, because with the help of intelligent agents, users can
browse the ontology that describes semantically the content of a set of data
sources containing pieces of software, and so they can select from it the
software (the service makes transparent for the users the location and access
method of various remote software repositories); and efficient, because agents
take care of reducing the wireless communication cost. This service is place
into the
ANTARCTICA project.
Current status
- Agents involved in the system
- Alfred functionality: it allows users to specify several ways to prune
the SoftOnt ontology: constraints or a level of detail.
- Software Manager: the SoftOnt ontology is built using categories and
programs in Tucows repository (off-line). The ontology is pruned by keyword or
by a level of detail (number of descendants strategy).
- Browser: it only allows users to navigate the pruned ontology. It
generates GUIs directly. Catalog refinements. Software selection.
- Catalog Updater: it only allows the Browser agent to update de catalog
that has been got from the Software Manager agent. This agent is created only if
the network connection has many errors.

- Constraints: the constraints are joined by an AND operator.
- The system estimate network performance.
- How many level of detail must be send to the user device.
- What is the strategy to prune the software ontology.
- Al data are send compressed to save network time.
- Test our prototype.
- The SRS is one of the finalist of the Cátedra
Telefónica challenge.
Future evolution
- Constraints languaje: Add new operator like AND, OR ...
- Measurement of performace and scalability.
- Test the system with mobile devices.
- Use a DAML+OIL build semiautomatically.
Other issues
The prototype has been implemented using
Voyager (mobile agents, CORBA),
Javacc ,
Jaxp and
JDK1.3 .
Instructions for demo in ra.
Screenshots