1 Overview

Searchy is a distributed metainformation search engine written in Java whose main goal is to federate different search systems. This federation provides a highly scalable easy to use system to locate resources within an organization's intranet or even in the Internet. In this context resource is used to mean anything with identity, it may be a data base entry, a simple file, a web page, etc.

The peculiar phylosophy of Searchy lets define it from many different points of view, you can select the one you feel better with. Searchy is ...

• A middleware layer that abstracts the search interface of several search systems.
• A federation of search systems.
• The middle tier of a generic three tier application with some extra functionalities.
• A gateway to translate local info to a general metainformation model.
• A web service implementing a complex semantic search engine.

Searchy takes a query and translates it into different search systems like data bases or directory services, and, in return, it expresses the result of the query in semantical terms. It is used to integrate different search services using a semantical perspective. An user that submits a query will see one service and semantical information, but in fact he will be accesing several different services, with different non semantic data formats.

A Searchy system is typically composed by several different agents working together. In small intranets only one agent may be enough, but if the intranet increases its complexity one might found necessary to use several different agents. Each agent may access to one o more data source (or data support). A data source might be any resource manager like a data base or a directory server. Actually three providers are supported: SQL data bases, LDAP services and the whole Internet, powered by the Google search engine.

Figure represents a quite simple example of use of Searchy. Here we can find a simple agent with two information providers accessing to a SQL data base manager and to a LDAP directory. Each agent is run in a machine, and, at last term, for the administrator it is only a traditional server like Apache.

Figure: Example of an agent with two providers

All the information that the system access is translated into an abstract semantic model, doing it we can manage information in the same way, regardless of its source and original format. The abstract model used is Dublin Core (DC) (see ), this model defines a set of fifteen terms that suit perfectly to describe resources. Mapping from the info fields avatible in the data sources to Dublin Core is one of the most difficult tasks of the administrator, read the map section for more details.

Searchy should not be accessed directly by final users but by an user interface, like a web application or a heavy GUI. Typically the final user will use Searchy thought a web form processed by a CGI, servlet o PHP page. Details about implementation of Searchy clients can be found in javadocs (see es.rediris.searchy.client.SearchyLocator), just as flash we can say that Searchy publish its service like web services, and any client may communicate with a Searchy agent using SOAP.