This paper reports a pragmatic experiment, consisting on the integration of a Web search engine with a traditional library's catalogue. PORBASE is the Portuguese bibliographic union catalogue. It holds more than one million of bibliographic and nearly half a million of authoritative structured records about authors, families, organizations and subjects. tumba! is a searchengine specialized in indexing web pages in Portuguese language. It indexes actually more than one million of pages, found potentially everywhere in the web. Since the early days of the Web that we have been assisting to a discussion about the pros and cons of web indexes built automatically by search engines as an alternative to traditional human-built databases, as the library's catalogues are. The main argument against search engines has been that they might be very effective in indexing the surface skins of the web, but they miss the richest contents hidden behind the more complex web sites and databases, the so called "deep web". On the other side their defenders point the relatively low cost of those solutions, which make it possible to provide good services without the costs of manual cataloguing, classification, indexing of the resources. This paper reportsabout an experiment to explore the best of both worlds, by using PORBASE in conjunction with tumba!. We extracted from PORBASE the major index of the authorities (names of persons), along with the frequency of each entry, and integrated it within the tumba! search engine. We also developed a very simple HTTP-based interface, ESPONJA, which tumba! can useto launch queries in PORBASE using the authoritative descriptions as arguments. In the scenario of this experiment, a user performing a search session using tumba! will be able to search also in PORBASE using the authoritative form of the search terms without knowing them in advance. The process starts when tumba! presents its own results complemented with suggestions (tips) to search in PORBASE. If the user chooses to launch a search there, a new window is opened showing the results of a search for the term. After that it is up to the user to continue interacting with PORBASE, to explore new results and take advantage of the specific functions of its specialized catalogue. This work aims to be only a proof of concept for this approach. Next steps will comprise the tuning of the algorithms to process the authority information in tumba!, and theformalization of the ESPONJA interface toward a stable, open, generic and reliable interface for interoperability with BN's specialized bibliographic systems.