Online Public Access Catalogs (OPACs) are complex systems that must respond to a multiplicity of requirements, often quite sophisticated. For example, they must be able to search for all of the works of a particular author, distinguishing them from those of other authors with the same name; or they must be able to search for a particular edition of a work; or different editions of the same work, even when they are published under different titles. Furthermore, they must deal with information of an administrative nature - for example, recording whether a book is available or out on loan. For this reason, the bibliographic information managed by an OPAC is augmented not only by a set of administrative data, but also by a complex infrastructure whose purpose is to permit the bibliographic control of the form of names, titles, editions, etc. It is in this way that OPACs distinguish themselves within the broader category of general information retrieval systems.

The international standard information retrieval protocol Z39.50 has been developed expressly to allow the user to avoid having to deal with problems associated not only with network connectivity, but also with different user interfaces, different names for elementary bibliographic items, and different languages. Increasingly we are seeing access to online public access catalogs via the World Wide Web proposed as an alternative to the Z39.50 protocol. But from a bibliographic point of view, there is no comparison between these two approaches. Although the Web might seem easier to use and capable of at least handling relatively simple searches, from a bibliographic viewpoint it is much less satisfactory, principally due to the underlying model of communication in the Web. In communication between client and server, HTTP plays a role of mere transport, and the browser plays an essentially passive role, limited to reconstructing locally the pages, forms, or applets that arrive from the server. As a consequence, the browser receives bibliographic descriptions in textual form, not as structured records which carry semantic information that can be utilized in other applications, and which allow visualization of the data in various forms that may be different from the form in which they are originally received from the server.

But the severest deficiencies become evident directly in the searching process. First of all, the problem of diverse user interfaces is not resolved. The consequences are perhaps not so evident in a national context, because the OPACs that are accessible via the Web within a single country tend to have fairly similar home pages, and thereby might give the impression that it could be possible to have a homogeneous form of dialogue. But even in national contexts, this impression is erroneous: although it is true that the “classic” OPAC access points are presented with easily identifiable names in all of the forms, it is also true that the access points of the individual OPACs may have very different access methods associated with them, and the user cannot be sure that the access methods he has in mind while creating his query (for example, access via string versus access via keyword) are really those associated with the access points he has chosen. This situation frequently results in search failure.

Z39.50, on the other hand, has been conceived to address directly the requirements of the user-system dialogue. Through the EXPLAIN facility and the SCAN service, Z39.50 allows the client to learn the characteristics of the server, the structure of its databases, the access points of each database, and the access methods associated with each access point. The application-level nature of Z39.50 enables the client to configure itself with all of the information known to the server and guide the user through the dialogue specified by the associated OPAC, yet always presenting the user with a uniform interface.

In summary, the Web can provide attractive solutions to many problems of information access and indexing, but the bibliographic domain requires the full-fledged application level capabilities that are embodied in the Z39.50 protocol.