Ontobroker applies Artificial Intelligence techniques to improve access to
heterogeneous, distributed and semi-structured information sources as they
are presented in the World Wide Web or organization-wide intranets. It reli
es on the use of ontologies to annotate web pages, formulate queries and de
rive answers. In this paper we will briefly sketch Ontobroker. Then we will
discuss its main shortcomings, i.e. we will share the lessons we learned f
rom our exercise. We will also show how On2broker overcomes these limitatio
ns. Most important is the separation of the query and inference engines and
the integration of new web standards like XML and RDF.