Z39.50 is an information retrieval protocol. It has generated much interest but is so far little deployed in UK systems and aervices. This article gives a functional overview of the protocol itself and the standards background, describes some European initiatives which make use of it, and outlines various issues to do with its future use and acceptance. It is argued that Z39.50 is a crucial building block of future distributed information systems but that it needs to be considered alongside other protocols and services to provide useful applications.
Perhaps it is its name - as if it had come from Krypton, so shorn is it of any clues about its meaning - that makes the significance of Z39.50 so hard to locate for librarians. On the one hand, it is seen as a panacea which will transform library and information services; on the other, it is dismissed as an irrelevance. Maybe if it were called the Information Retrieval Protocol, along the lines of the File Transfer Protocol, it would seem less strange and more easily assimilable. As of June 1995, the Z39.50 Register of Implementors maintained by the Z39.50 Maintenance Agency at the Library of Congress listed over 80 implementors. Yet Z39.50 is not yet widely deployed in the UK: there are no routine production services which make use of it. It is not yet widely deployed in the US or Europe either, but there seems to be rather more development effort, and applications are closer to deployment in real services.
We will say something about how Z39.50 works in Section 2, but our main purpose in this article is to give a better sense of what it is and what its implications are by describing some of the projects and services currently making use of it. We concentrate on European developments as these are rather less well covered in the literature than transatlantic counterparts. These are quite various in scope and ambition. In some cases (Pica, IRIS), the protocol is being used in support of operational, or near-operational, services; in others the focus may be more on software development. We have not aimed to be comprehensive, and some projects are not described in great detail. We then examine some general issues, including the relationship between Z39.50 and the Web; the availability of Z39.50 products in the European marketplace and describe relevant interested organisations.
Z39.50 is a retrieval protocol which allows client applications to query databases on remote servers, to retrieve results, and to carry out some other typical retrieval-related functions. Z39.50-1995, or Version 3 of the protocol, has just been approved by ANSI, the American National Standards Institute. This enhances Z39.50-1992, or Version 2 as described in 2.3. Most current implementations are of Version 2. An early version of the protocol, Z39.50-1988, was popularised as the communication protocol in WAIS, though this does not interoperate with later versions. Z39.50 allows application builders to separate the user interface from database applications, and provides a means for the two to communicate. It uses a client/server model. (In Z39.50 terms, the client is called an 'origin', the server a 'target'.) The client and server communicate via the Z39.50 protocol, as shown schematically in Figure 1.
The protocol provides generalised facilities for the communication of queries and results. One important feature is the attribute set. This allows terms and various of their characteristics to be defined. They capture the semantics of a particular area. For example, the attribute set BIB-1 includes a range of elements typically found in bibliographic applications. Z39.50 also uses a generalised search syntax. Together these allow the syntax and semantics of bibliographic type queries to be expressed independently of any particular search system or data format. The user application will convert queries accepted at the interface into the form required by Z39.50; these will be communicated by the client to the server. At the server end, another application will take these and convert them into an appropriate form for running against databases. Results are communicated as records; a number of record formats are acknowledged: MARC records, and some newly defined Z39.50 specific formats. For example, the OPAC record is a structure defined to communicate bibliographic, holdings and circulation data. Records can be taken by the user application for further processing. They may be presented at a user interface or passed to another application.
Z39.50 is a program to program protocol: it is not something an end user sees, or, ideally, is even aware of. A user interacts with a user interface: it might be part of a library system, it might be part of a standalone Z39.50 client application. The rather loose use of the word `client' should be noted here. In common currency, `client' is often used to cover all the components presented on the `client side' of Figure 1. However strictly speaking, the client program is that component which implements the protocol. Typically this will present an API (Application Programming Interface) to which applications are written. These may be stand-alone applications as is the case with the several Windows based `clients' that are now marketed, or they may be embedded in another system. One of the aims of the SOCKER project as described in 4.5 is the implementation of a client program in three different `client-side' application environments.
The user is shielded from search engine or display differences. Because communication between client and server is standardised, the client can interoperate openly with Z39.50 servers over the network. The behind the scenes transformations mean that users have a consistent view of a variety of databases through the same user interface; they do not see beyond their own user interface.
Thirdly, the protocol allows the actual retrieval of structured records. It thus allows `smart' applications to be built based on a shared understanding of this structure. So, for example, user applications will be able to consolidate data from multiple sources, do format conversions, as well as formatting for display. Data can be passed on to other applications (a request service for example), or used to trigger other actions (one can imagine how an automated Selective Dissemination of Information service might be created).
A Z39.50 session consists of an association between an origin and a target. Protocol Data Units (PDUs) are the means of communication between the origin and the target. Nearly all PDUs work in pairs; in most cases, the origin sends a PDU to the target, which replies with another PDU. The former of these is usually some request, such as a search request or a request to delete search results, while the latter is usually a response, such as details of the search or confirmation of search result deletion. To create an association a Z39.50 association is opened by the passing of an InitialiseRequest PDU from the origin to the target, and the reply of an InitialiseResponse PDU in the opposite direction. This exchange of PDUs establishes whether the origin and the target can operate together, by defining parameters acceptable to both such as the version of Z39.50 that either can support. The result of this exchange is that either an association is made, or that some incompatibility between the facilities that the origin/target can provide means that the connection is terminated.
PDUs are grouped into several services, which are defined by Z39.50; the most important of these are the search service and the present service:
Initialisation service - deals with trying to set up a successful association between an origin and a target.
Search service - deals with search requests. The SearchRequest PDU, sent from the origin to the target, contains the user's query, formatted into a Z39.50 compliant state. The target takes this query, carries out a search on the database, and stores the results. It then sends a SearchResponse PDU back to the origin, giving details of the search results. Note that the search results themselves are not transmitted to the origin, only details about them. Some targets can store the results of many searches at any one time.
Present Service - here, the origin can request sets of results. The PresentRequest PDU, paired with the PresentResponse PDU, is used to specify that some set or subset of results from a previous search should be sent to the origin by the target.
Access Control - this service covers security measures. The AccessControlRequest PDU is sent by the target, whenever it wants, to the origin, which must respond with a valid AccessControlResponse PDU before any other operations are carried out within the current association.
Resource Control - this service covers the use of resources within the current association. The ResourceControlRequest PDU is sent by the target to the origin, containing details of resource allocations required by the most recently requested origin-based request. The origin decides on whether the request is too resource-intensive, and sends a ResourceControlResponse PDU back, telling the target to either carry on with the operation or to terminate it.
Delete Service - here, the origin can request that the target deletes one or more sets of results by use of the DeleteRequest/Delete Response PDUs.
These services are part of Z39.50-1992. It should be noted that Z39.50-1995 enhances and extends these services, clustering related services in several facilities. Important additional services are provided which better equip Z39.50 for use in real world applications. Services include Explain and Extended Services. The former allows the client to retrieve information about the server-side application and databases: for example, the databases up for searching, restrictions, availability and so on. Explain is not yet widely implemented. Extended services is an important enhancement. It allows various services to be requested which are carried out by the server-side application outside of the Z39.50 session. They are not Z39.50 services, but additional services identified as requirements within the various contexts in which Z39.50 will be applied. Extended Services so far defined include saving a result set for later use, updating a database and, importantly, item order [1]. In each case information may need to be retained outside of the Z39.50 association in which the service was requested.
The rapid growth of diverse electronic information resources in a computationally intense networked environment continues apace. The Internet now represents a massive aggregate of heterogeneous, autonomously managed information resources. Without automated support for the discovery, location and use of information resources, the user will be overwhelmed by the volume and variety of available resources. The flows of information, the creation of communicating applications, and the construction of integrated end-user environments will benefit from the deployment of resource discovery systems, a constrained set of search and retrieve protocols, agreed ways of structuring and formatting data for exchange and processing, and a protocol framework for charging [2]. We are currently putting in place a number of building blocks. Z39.50-based services will be an important building block.
Z39.50 will allow bibliographic applications to be made available through standard server interfaces thereby facilitating distributed applications, the integration of bibliographic databases into wider networked information systems [3], and the direct application to application exchange of structured data. Currently bibliographic resources are available for terminal access, with a variety of different interfaces. Bibliographic systems are insulated from each other: they are not true network applications.
One factor in the delayed adoption of Z39.50 has been a double confusion: at the application layer between Z39.50 and SR (Search and Retrieve) and, in lower layers, between TCP/IP and OSI.
Z39.50 is an American national standard, which originally emerged in the context of the Linked Systems Project in North America. Search and Retrieve is a parallel ISO standard, influenced by early Norwegian and other work. (This historical development is described more fully elsewhere [4].) Z39.50-1992 aligned itself with SR, with minor differences and the addition of some extra services. However the critical mass of developers, the openness and responsiveness of the standards making process and the appearance of working implementations has meant that the development of Z39.50 has outstripped that of SR. In fact, SR has not been widely implemented. There may be one or two SR servers in existence, but there is none in routine production use. SR development effort has lacked the concentrating focus of the US Z39.50 Implementors' Group: the ISO standards process is slow, and there are very few products or services which implement SR. For these reasons, it has been suggested that the two protocols converge, and at the time of writing, a decision is about to be taken as to whether SR adopts Z39.50-1995 in its next version.
SR and Z39.50 are defined as OSI application layer protocols: they expect OSI lower layer services. For pragmatic reasons, most Z39.50 developers have chosen to implement Z39.50 over TCP/IP and to do away with OSI services. Given the importance of the Internet this is a natural decision. Another factor is that existing implementations of the OSI stack are large and do not facilitate the development of PC-based applications. Accordingly the development effort has been split between the TCP/IP and OSI environments: applications created in these environments do not interoperate. The EUROPAGATE project, described in 4.3, was conceived in this context, and aimed to build relevant interworking solutions. A couple of years ago, it was reasonable to suggest that Z39.50 over TCP/IP characterised the US environment, and SR over OSI characterised the European. This is no longer the case. Z39.50 over TCP/IP is the dominant applications environment for pragmatic reasons. Indeed, one might suggest that were it not for the European Union support for OSI and SR, that such SR development as there is would be much diminished. (It should be noted that Z39.50 over OSI has also been implemented, but it is very rare.) At the same time, there are signs that interest in OSI may be reemerging as it is acknowledged that some presentation layer and other services offer advantages, as mOSI (minimal OSI) and other approaches open the way for light-weight OSI applications and further work facilitates smoother interworking with TCP/IP. However, at this stage, it is unclear how important these developments are likely to be.
IRIS is interesting as it represents one of the first operational Z39.50 based services. The development work and the resultant service are fully described elsewhere [5]. IRIS now operates a commercial document awareness and supply service aimed at business, libraries and other intermediary services in the Republic of Ireland. It allows users to search for and request items of interest from the catalogues of six Irish libraries, and from the UnCover service. The initial impetus came from a desire to more effectively share resources between Irish libraries, and to make those resources more widely available. A strategic decision was taken to create a consolidated resource from the transparent linking of independent library catalogues rather than invest in the creation of a union catalogue. Support was sought from the Telematique programme, funded under the European Regional Development Fund, to implement a proprietary gateway system, the Irving Library Network, between the libraries participating in the scheme. The application was successful and the IRIS project was set up in 1992 with a budget of ĢIR600,000, half of which was met from participants' own funds. After contact with UKOLN in 1992, the project emphasis shifted and it was decided to adopt a standards based approach, and to use a Z39.50 based application as the core component of the service. The Telematique funding also meant that what might have otherwise have developed as a library resource sharing project now has a strong end-user oriented thrust, and a major target audience in the business sector.
Six libraries agreed to proceed on the basis of a feasibility study into technical and business options carried out by Euristix Ltd. in Autumn 1992: Dublin City University (DCU), Trinity College Dublin (TCD), University College Dublin (UCD), University College Galway (UCG), University of Limerick (UL) and Forbairt, the national science and technology research agency. At the time of writing the IRIS service agency, based at Forbairt, presents the service on behalf of IRIS Document Delivery Services, Ltd., a limited company owned by the IRIS participant institutions. Different service arrangements exist for different categories of user: participant library users have different terms and conditions than business users, for example. In a parallel development, IRIS libraries report their holdings to UnCover, so that IRIS library users can identify items of interest available through the IRIS system. IRIS libraries have also agreed to become suppliers of documents to UnCover users, and UnCover staff are based in the libraries to fulfil requests.
The IRIS system architecture is presented in Figure 2. All users initially connect to the IRIS client host, which offers an integrated range of services necessary to deliver a commercial service: searching services, requesting services, management and accounting services. Search requests are converted to Z39.50 requests and passed by a Z39.50 client to a target library. A particular feature of IRIS is that it caters for several Z39.50 connections to be opened simultaneously. A user can select to search one or more libraries by toggling them on and off at the user interface. The one search is run against all selected databases, and the IRIS system collates returned results and presents them meaningfully to the user. The target libraries have implemented Z39.50 server capabilities in association with their library systems (Urica, BLCMP, Dynix, and Oracle library systems are in use). However, currently, UnCover is not available in this way -terminal access to the service is provided. The request service is E-mail based and uses some of the data elements defined in the ILL protocol, though not the protocol itself.
The host client system was developed by Fretwell-Downing. It implements Z39.50 version 2 over TCP/IP. Users have terminal access to a character-based interface on the client host. A windows-based interface has now also been developed which communicates with the client over TCP/IP. Only one host system is currently in operation, though in principle others may be put in place. Effectively what it provides is a unified approach to the management of a distributed resource, and could be deployed in a number of operational environments. Fretwell-Downing are using some of the experience gained here in the DALI project. There is also an overlap in participation between these projects and EUROPAGATE.
DALI (Document and Library Integration) is a project funded under the European Libraries Programme. It began in early 1995. The project will "develop, pilot and evaluate a service for multimedia document delivery in a distributed environment, using SR" [6]. The co-ordinating partner is Fretwell-Downing Ltd. Other partners are University College Dublin, Institute of Oceanographic Science, Sheffield University, Blackwell Ltd (UK); Kyros, Thessaloniki University Library (Greece).
The focus of the project is on the development and proving of a unified service architecture based on standard components. The aim is to deliver an integrated service for search, request and delivery of items to the end user, which hides the mechanics and complexity of the interacting systems of service provision. Oceanographic studies has been selected as an initial test domain. There are three main components: a DALI client, database servers, and the DALI server [7]. A light-weight DALI client manages interaction with the user. Resources are made available through database servers: these might be bibliographic resources, document servers, and so on.
The interesting part is the slightly confusingly named DALI server which is responsible for managing the set of services for searching databases, requesting documents, and delivering them. The DALI client and DALI server communicate using a proprietary protocol. The DALI server communicates with database servers using standard protocols (WAIS and Z39.50 for search; ILL over SMTP, http, Z39.50 item order for request; fax, E-mail, and http for delivery). It also provides system management and accounting services, as well as user authentication and registration services, which allows the server administrator to determine individual user privileges. The DALI server also adds value in other ways - through collation of results, format conversion and so on. DALI can be seen as an enrichment of the IRIS service environment: IRIS provides search, request and management services; DALI enhances and extends these, and adds delivery capabilities. Each aims to provide a systems environment which allows a distributed resource to be managed effectively for end user access. They provide an interesting contrast to the rather naive view that all that will be required is a Web or Z39.50 client on every desk. They provide several services and some necessary integrative glue.
EUROPAGATE is another project funded under the European Libraries Programme. The partners are University College Dublin, an Chomhairle Leabharlanna (Irish Library Council), the Danish Technical Knowledge Centre and Library, and the Consejo Superior de Investigaciones Cientificas, Spain. The project began in early 1994, and at the time of writing is about 70 per cent complete. The technical aims of EUROPAGATE can be summarised under the following headings:
The factors in the communications environment which prompted this project are described in Section 3 above. However, during the course of the project changes in this environment suggest a shift of emphasis down this list. The need for (i) has receded since the proposal was accepted, as the protocols converge; (ii) remains an issue, although very few servers are currently implemented in an OSI environment. Some will arrive (maybe DBV OSI, as described in 4.8), but in the current operational environment there is little need for such a gateway. This is not to deny the potential value of this part of the project, given the possible future emergence of more servers in a commercial (and possibly non-Internet) environment and the potential importance of mOSI (minimal OSI) in facilitating more OSI applications. Although, at this stage, what use of it will be made is really unknown. Under (iii), the gateway provides support for conversion between different MARC formats, some administrative and management support, and the ability to submit queries and receive results from the gateway via E-mail. In the latter stages of the project, support for Web access will be provided. http is a stateless, or connectionless, protocol: it preserves no information between accesses. Z39.50 is a stateful, or connection-oriented, protocol: it retains information over a user session. While building the E-mail gateway, the developers have added an extra software layer that preserves this information transparent to the client. This will now also facilitate the construction of the http to Z39.50 gateway, and allow it to be more functional than some of the other such gateways currently available. The developers have constructed the gateway from existing software components, in particular Nordic SR, IRTool from the National Library of Canada, and Zdist from CNIDR. The project found that more effort than anticipated was required: they had to fix bugs in everything used, and support and documentation was limited. The gateway includes a Z39.50 server running over TCP/IP and an SR one running over OSI. It will accept queries from Z39.50 clients over TCP/IP, SR clients over OSI, and E-mail (queries need to be formatted in the Common Command Language) and will run those queries against servers in either environment. The project has demonstrateda working system which can be accessed by existing standard clients, and, which, they claim, adds little delay in the search process. In the absence of the Explain service, which provides information about servers and databases, the gateway has to be configured with server data.
The project expects to make the gateway publicly available. They have identified and described a number of service environments in which it might be used.
Pica is a Dutch organisation which provides automation, union catalogue and information services on a national level to academic and other libraries. Recently it has extended its operations into Germany, and the Pica shared cataloguing and library automation infrastructure is being replicated in several German länder.
Pica is pioneering the use of network applications by libraries and they are involved in a range of national and European complementary initiatives. The size of The Netherlands, good support for library and information services, and their nationwide presence have benefited Pica. The focus they provide supports a co-ordinated view of developing information services and greater attention to infrastructural issues at a national level than is the case, for example, in the UK.
They have put in place the infrastructure to support systems and services necessary to realise an integrated system for resource sharing and are now emphasising the importance of end-user information services. They have, or will soon have, services which allow a user of a member library to search a range of resources (local catalogue, catalogues of other libraries, the national union catalogue, a table of contents service, abstracting and indexing services, remote OCLC services), to request documents (through their ILL system) and to have documents delivered electronically and otherwise.
Pica was early to recognise that future distributed library and information services will be based on standard protocol-based approaches: they have played a leading role in the development of the library-oriented OSI application layer protocol standards ILL (ISO 10160/1) and Search and Retrieve (ISO 10162/63), ILL especially, and they are now active contributors to the Z39.50 Implementors' Group. They are carrying forward their experience into project ONE and are partners in DBV-OSI II. They have also been active in developing the GEDI framework for electronic document delivery (they have a national project called RAPDOC in this area and are working with the Research Libraries Group in enhancing the Ariel workstation). They were a partner in Project ION (Interlending OSI Network), now complete. They are also involved in the European projects EDIL and Edilibe II which are looking at standard approaches to electronic document delivery and EDI respectively. Recently, in collaboration with libraries in The Netherlands and Germany and some international publishers, they have introduced a new web-based service which provides access to collections of documents in various formats and a central catalogue of what is available.
These developments have been driven by service goals: to provide seamless service interfaces to cataloguing and information resources. They are implementing standards in service environments as they become stable. For example, the Open Library Network (a project aiming at linking the library OPACs) was not initially based on Z39.50: it is currently being migrated to Z39.50. At the same time selected libraries are using the OCLC product SiteSearch to provide integrative access to OCLC and Pica resources through Z39.50 interfaces. They have a number of other Z39.50-based products and services. For example, cataloguers use a Z39.50 link to RLIN services. Pica has pioneered the use of operational Z39.50 services and is collaborating in DBV-OSI II and Project ONE, placing it at the heart of European Z39.50 development. Z39.50 is one strand of its distributed service environment.
The SOCKER project [12] resulted from the European Libraries Programme's first Call for Proposals; it started in December 1992 and is scheduled to finish at the end of 1995. It is co-ordinated (like PARAGON - see below) by UNI-C, the Danish Computing Centre for Research and Education; there is one other Danish partner (DBC), as well as IME (producers of the TINlib library management system) in the UK.
The central part of the project (which has now been completed), was to produce an SR client toolkit. This enables developers either to define new client applications, or to integrate the code into existing ones. Each of the project participants is using the toolkit to develop an application as part of the project:
The SOCKER kernel has been developed so that it will function either over the full OSI stack, or directly over TCP/IP. Like Nordic SR-Net, the latter option was added mainly to allow interoperation with Z39.50 implementations in North America over the Internet. The project itself has used ISODE (ISO Development Environment). The lack of SR targets available for test purposes will come as no surprise: out of the two initially identified (Pica's target for Project ION and DBC's Danbib), the Pica product is no longer available because of the ending of ION. Therefore the SOCKER software has also been tested against several Z39.50 version 2 (some with parts of version 3) targets in the US; this was reported to be successful. Several limitations of the SR protocol (such as the lack of index scanning and proximity searching) became apparent during the course of the project. However, although both the above services are included in version 3 of Z39.50, it was decided to remain with the SR standard for the purposes of the project.
It is worth mentioning in this context Yaz, a public-domain Z39.50/SR toolkit which was developed by Index Data in Denmark [13]. It supports both client and server implementations and results from the company's development work for the SOCKER and EUROPAGATE projects.
As a precursor to the ONE project, it is worth looking briefly at Nordic SR-Net [14]. The project started in 1991 (so was relatively early) and was completed in 1994. As its name suggests, it involved the five Nordic countries, with Iceland as an observer. Funding came from NORDINFO (the Nordic Council for Scientific Information), the national library authorities and the participating organisations themselves. It is important as one of the very few functioning SR implementations, allowing communication between five shared and union catalogues running on different systems. It is interesting that although the main implementation was SR over ISODE, the project decided to add the capability of running SR directly over TCP/IP. This was done mainly so that they could interoperate with North America Z39.50 implementations, which usually run directly over TCP/IP. However, it transpired that there was actually a serious error in the SR Net software, which meant that the project partners could only interoperate between themselves, and not with other SR implementations. But as there are no real working SR servers around, this is of little consequence! Instead, the attention of the main players in Nordic SR-Net has switched to the ONE project.
ONE is an important project, resulting from the European Libraries Programme third Call for Proposals [15]. It started officially at the beginning of 1995 and it probably involves the largest number of participants and countries yet: 15 and 8 respectively. This was in fact because two project proposals were merged to form one eventual project. Like Nordic SR-Net, the co ordinator is BRODD, the R&D and consultancy department of the Norwegian School of library and information science. However, partners and associates in ONE are far more wide ranging than the earlier project: the British Library, Satellites International Ltd (UK); Danish Library Centre, National Museum (Denmark); Die Deutsche Bibliothek; Pica (The Netherlands); Joanneum Research, Die Steierm=E4rkische Landesbibliothek, Steierm=E4rkische Landesmuseum Joanneum (Austria); LIBRIS (Sweden); the National Library, BIBSYS, University Library of Oslo (Norway); Helsinki University Library - TKAY. It is therefore significant not least in its encompassing of a number of major European institutions. The project will interconnect the OPACs of the participating national libraries and cataloguing services and will offer a trial service to access any of the catalogues via a single entry point. Most participants are already developing services based on Z39.50 or SR, so they plan to use their existing systems as a basis for further development and subsequent integration of these systems. The British Library, like some of the other players, intends that future service implementations of Z39.50 will be based on products from ONE. One of the key aims of the project is to produce origin and target public domain software which is able to run over both Internet and OSI protocol stacks. The issue of character sets will also be addressed. An expected result of the project is the establishment of a base for a pan European OPAC network. ONE will be using the Z39.50 software which was developed by Satellites International for the German DBV-OSI II project.
Unlike most of the other projects being covered here, DBV-OSI II is not a project funded by the European Commission. It is a German national project, with federal funding [16, 17]. There are eight partners. The co-ordinator is Die Deutsche Bibliothek, with Danet providing technical project management. Other partners are the online hosts STN and DIMDI, and the regional union catalogue organisations Deutsches Bibliotheksinstitut (DBI), Bibliotheksverbund Bayern (BVB), Bibliotheksverbund
Niedersachsen/Sachsen-Anhalt/Thüringen (BRZN) and Südwestdeutscher Bibliotheksverbund (SWB). The final participant is the Dutch company Pica -several of the union catalogues involved are automated using the Pica library system.
The idea is to provide an integrated infrastructure for searching, provision of holdings information, requesting and delivery of materials, all based on Z39.50. Users should then have transparent access to a consolidated resource using the familiar local interface. The project is important as one of the few concerted efforts to date to tackle the problem of linking the three areas of distributed searching, access to holdings information and requesting. We are currently interested in this area at UKOLN and will be addressing some of the issues in our MODELS project (see below).
Satellites International Ltd (SIL) in the UK were responsible for developing software for the project, providing a subset of Z39.50 version 3 services. The API supports TCP/IP, but can also use ISODE (ISO Development Environment). A specific technical approach was chosen to ensure maximum portability and reusability of the software developed for the project. It is therefore interesting that the software will also be used by the ONE project as described in 4.7. A Microsoft Windows version of the client software is being developed.
Interoperability testing is due to start at the end of September and is expected to take some time, given the eight participants. Beta testing is scheduled to begin with a workshop at the beginning of 1996. A group of chosen libraries and information centres will then take part in the pilot phase. In phase two of the project, which will commence in 1997, it is also planned to implement the ILL protocol with support for document delivery. The project is also considering the use of Z39.50 item order extended service for requesting of materials.
ARCA was selected as a funded project under the European Libraries Programme third Call for Proposals [18]. In contrast to some of the other projects, ARCA involves just two southern European countries: Italy and Spain. The co-ordinator is Intecs Sistemi; other Italian participants are CNR.IEI, Regione Toscana - Servizio Beni Librari and Università degli Studi di Pisa. Spanish participants are Sabini Automatizacion de Bibliotecas and Fundacion Sancho el Sabio.
ARCA is to develop an easy-to-use software tool which will allow existing OPACs to act as SR targets. This will be made generally available. Depending on the structure of the local library software (e.g. presence of an API), it will still be necessary for libraries to contribute varying amounts of programming effort to write code specific to their OPAC. The tool will be implemented for two library systems: CDS/ISIS in Italy and SABINI in Spain. These are two of the most widely used packages in their respective countries. Other systems being used by Italian participants (ATLAS, ERASMO and UNIBIBLIO) were used in carrying out the OPAC functional analysis which formed part of the background work to define the user interface requirements.
A key principle of the project when originally designed was the so-called `dictionary', which would store the particular characteristics of individual OPACs. This would mean that a common SR kernel could be used to connect several different types of OPAC. It may also be used on the client side, in order to filter non-relevant requests. However since the architecture was designed, version 3 of Z39.50 has been accepted. The complicating factor is that version 3 incorporates the Explain service, which bears some resemblance to the project's dictionary concept. The project has said that subject to further investigation, it might be possible to combine both.
The project will also develop an SR client with a graphical user interface. A particularly interesting feature is that it will include functionality to define search domains. These are sets of target libraries which can be searched simultaneously.
PARAGON also started at the beginning of 1995 and builds on the work of an earlier project called JUKE-BOX. Like SOCKER, it is coordinated by UNI-C in Denmark, but this time is looking at the problems of creating SR servers for sound archives. Other partners are the Danish Statsbiblioteket and the UK and Italian National Sound Archives. Initially the project will develop access to the standard catalogues of both these archives. However, later (and clearly this will be the most interesting bit) it intends to investigate the transfer of sound as well.
An important part of the project will be to investigate the implementation of SR on top of the `slimmer' OSI stacks mOSI (minimal OSI) and tOSI (thin OSI), which offer (in relative terms) a simpler development environment.
EURILIA was selected as a result of the second Call for Proposals for the European Libraries Programme and so commenced at the beginning of 1994 [19]. It is co-ordinated by the University of Limerick in Ireland and involves Cranfield University in the UK; the Technical University of Delft and Digital Equipment Corporation in The Netherlands; Sup'Aero, ENSEA in =46rance; and Instituto Nacional de Tecnica Aerospacial in Spain. Four of th= e library partners have major aerospace collections. The central aim of the project is to develop a standard interface based on SR which will provide common access to the OPACs of aerospace library collections. This will be demonstrated between the participating libraries as part of the project. =46acilities for image browsing will be provided (a pilot image database has already been set up), and document delivery will be available.
The recent success of the Web has prompted some to query the need for Z39.50. This suggests a lack of understanding of what Z39.50 might be able to support in the future and is to misconstrue the relationship between the approaches. In a sense, the Web provides added value terminal and file transfer abilities. It provides a unified terminal interface to a navigable resource space. Gateways to non http (the Web protocol) resources can be created. However, for most current library applications the user is typically dropped into a telnet session. A growing number of bibliographic services are being made available through cgi (common gateway interface) interfaces. A gateway is written to a server, and the user is offered forms-based access. Where all that is required is access to a particular resource, this makes sense. However the work has to be repeated for each resource. As a next step, a http-Z39.50 gateway allows a number of resources to be made available through the same interface, and allows a user to access the Z39.50 `information space' using their familiar Web browsers. However, the structured output from the bibliographic database will typically be converted to a HTML document for display; the structure of the returned records is thrown away because the Web has no use for it. One can imagine an application built on Z39.50 which is smart about the returned data because it understands its structure. It could tag it for input into a personal bibliography; take a record and search in another database for works by the same author; pass the record to a document request service; and so on. Z39.50 allows the creation of interworking bibliographic applications which share structured data. Of course the eventual output of such a service might be delivered to the user through a Web interface. The Web begins where the need for smartness ends. Thus the Web will probably be used to provide a user interface to particular resources or applications. For example, a user of the services provided by EUROPAGATE or DALI might access them using their Web browser. But where interworking bibliographic applications are required then there is currently no real alternative to Z39.50.
Many of the publicly-funded projects described in Section 4 state that the Z39.50/SR software developed will be freely available at the end of the project. However, these are mostly toolkits intended for local development of specific applications; in addition, many are several years away from completion. The majority of libraries will not have the inclination or indeed the need to carry out their own Z39.50 development work. In terms of Z39.50 clients and servers, there is now a fair number of off the-shelf products available on the marketplace, even in Europe [20].
The commercial clients and servers are being marketed by the traditional library management system suppliers. Both those with full working clients and/or servers and those with prototypes are listed in Table 1. Clients can either be fully integrated with the standard OPAC, or available as a separate stand-alone product. The integrated clients have the benefit of offering the familiar user interface for local and remote searching, a precept of the Z39.50 protocol. Stand-alone clients could feasibly have the same interface, but like Dynix's WinPAC, tend to be different to the standard OPAC. This therefore disregards a key advantage of Z39.50. However being stand-alone means that libraries (or even other organisations) whose own system supplier is not yet offering a Z39.50 client, still have the option of purchasing one. Geac's stand-alone Z39.50 client, GeoPac, is designed as a Windows interface for searching both local and remote databases. This would normally assume the existence of a local Z39.50 server, although GeoPac itself has facilities to use the internal Geac stand ard as an alternative if there is no local Z39.50 server. It is unsurprising that the majority of systems with Z39.50 compatibility are those marketed in the academic/larger library sector. One exception is TINLIB; this is explained by IME's participation in the European SOCKER project as described in 4.5. Z39.50 implementation is a long and costly business in research and development terms. The cost and effort factor required on the part of libraries is also a reason for the acute shortage of Z39.50/SR servers actually up and running in Europe.
As libraries offer an ever expanding range of services, for end users in particular, it is unlikely that a single supplier will in future be able to cater for all their needs [21]. So mixing and matching of modules and services from various sources is likely to become increasingly common. This again reinforces the need for an application protocol such as Z39.50 to link the parts together.
EFILA was set up at the beginning of 1995, by EWOS EG-LIB in collaboration with DG XIII/E-3, Library networks and services. Its aim is to provide a forum where (mainly) implementors can discuss issues of common concern relating to SR/Z39.50, electronic document delivery, etc. Participants are largely those already involved in European Commission funded projects. There are two meetings a year, held in Brussels so far.
A significant step was taken at the last meeting in June 1995 when it was decided that EFILA would co-ordinate a European input into the ZIG protocol development. This recognises the need for an increased European contribution to the international standards making process.
The ZIG was formed in 1990 and is responsible for the technical development of the Z39.50 standard. It has an extremely active mailing list, where suggestions for modifications and enhancements are posted and then discussed in detail. As Z39.50 is a US national standard, ZIG meetings (two to three a year) are normally held in the US. However, in recognition of an increasingly active body of European implementors, the first European ZIG meeting was held in Amsterdam in April 1995. The September 1996 meeting will also be in Europe (Brussels) and it appears that one European meeting per year could become the norm.
Co-ordination and organisational support for the standard is provided by the Z39.50 Maintenance Agency; the Library of Congress fulfils this role. Among its tasks is the compilation of the very useful register of Z39.50 implementors.
The UK Z39.50 Pre-Implementors Group (PIG) was set up following a UKOLN workshop in 1992 with the aim of encouraging UK involvement in the development of search and retrieve standards and contributing to international technical discussions and ballots. It was renamed the UK-ZIG in 1995, to reflect the fact that a number of UK organisations have now succeeded in implementing the protocol. Participants come from a variety of backgrounds including suppliers, libraries, UKOLN, the British Library and LASER. As a forum for exchange of ideas, implementors have found the group useful. The group also owns a mailing list [22].
Finally we turn to UKOLN's MODELS project (formerly named Project Distribute). It arises from a perception that there is a general lack of understanding about what distributed library services will look like, how they will be integrated into the wider network environment and what steps now need to be taken to move this forward. Although there has been discussion of Z39.50 and other protocols, there is no shared view of how they will be used to support distributed services. An applications framework which enables the effective management of distributed collections of resources and services now needs to be constructed. MODELS will provide a forum within which the UK library and information community can explore some of the issues that need to be addressed. It is being supported by the UK Higher Education Electronic Libraries Programme (eLib [23]). In order to make the project more manageable, it has been divided into five project lines:
Each of these will be explored as individual, although linked, studies which will last for approximately five months. There will be some parallel running, with the overall project to be completed in 18 months (from a start date in August 1995). The programme of work for each line will include:
Although devised and led by UKOLN, the project is working closely with other groups such as eLib, the British Library, Project EARL and UK-ZIG. This will ensure widespread community involvement, relevance to emerging developments and impact. A Steering Committee has also been appointed. In addition, each project line will also have an individual advisory panel, which will advise on direction and help to plan the content of the workshops. We are currently working on the background review for distributed document discovery and supply and plans are underway for the first workshop in December 1995.
One of the main objectives of the project is to raise awareness of the issues within the library and information communities. We will, therefore, be producing a range of project papers and reports, together with appropriate Web pages and items in the Ariadne newsletter [24]. (UKOLN is also developing a UK/European Z39.50 homepage.) [25]
Therefore to use the language of FIGIT (Follet Implementation Group for IT) circles, MODELS is addressing the `plumbing' aspects or `Follett glue'. These curious terms have been used to refer to the systems framework which will hold together the variety of services which will comprise the electronic library of the future. It is likely that Z39.50 will play a major role in providing some of this necessary plumbing.
Z39.50 is not an end in itself. It is one building block in bringing library and bibliographic services more fully onto the network. It is apparent that there is quite a lot of European activity. There is rather more in the US, which we have not reported here. One can identify several strands of activity, and points of contact. DALI, EUROPAGATE, and IRIS have some shared partners, and this partnership may be continued in further projects funded in the libraries section of the European Union's Fourth =46ramework Programme. (The results of the first call have not been made public at the time of writing.) Another major axis of development is represented in Project one, which brings together some of the experiences of Pica, DBV-OSI II and Nordic SR-Net. Most of the projects are in research and development stage; some have moved into production. There is now a variety of client and server products on the market. Major organisations such as Pica and OCLC have made it a central part of their strategic service and technical goals. Z39.50 is clearly going to be a central: it will be no panacea but it cannot be ignored.
[1] Fay Turner. Document ordering standards: the ILL protocol and Z39.50 Item Order. Library Hi Tech, vol. 13, no. 3, 1995, pp. 25-38.
[2] Lorcan Dempsey, Anne Mumford and Bill Tuck. Standards of relevance to networked library services. In: Libraries and IT: working papers of the Information Technology Sub-committee of the HEFCs' Libraries Review. Bath: UKOLN, 1993.
[3] Lorcan Dempsey. Network resource discovery: a European library perspective. In: Neil Smith (ed) Libraries, networks and Europe: a European networking study. London: BLRDD, 1994. (LIR Series; 101).
[4] Lorcan Dempsey. Libraries, networks and OSI. Westport CT: Meckler, 1992.
[5] Annette Kelly and Bryan Alton. IRIS: a Z39.50 based service for database searching and document ordering. Journal of Information Networking, vol. 2, no. 3, 1995, pp. 187-213.
[6] Synopses of projects (Release CfP'91, CfP'92 and CfP'93). Luxembourg: European Commission, Directorate General XIII, 1994. (For further information about the Libraries Programme contact the Library networks and services unit - Fax: +352 4301 33530)
[7] Details based on Synopses document and personal communications.
[8] Description based on project reports available at <URL: http://www.dtv.dk/egate/egate.html> and on personal communications from project members, 27 July 1995.
[9] OBN final report: from project to library users / Stichting Centrum voor Bibliotheekautomatisering Pica and Surfnet. - Leiden/Utrecht: Pica/ SURFnet, 1992.
[10] Pica: progress report April 1994 - June 1995. Paper distributed at European Library Automation Group meeting, Trondheim, June 1995. Dated 12th June, 1995.
[11] Look Costers. A managed information network. In: Lorcan Dempsey, Derek Law and Ian Mowat (eds) Networks and the future of libraries 2: managing the intellectual record. London: Library Association Publishing, 1995.
[12] Erik Bertelsen (ed.). SOCKER first edited report. May 1995.
[13] Yaz can be downloaded from <URL:ftp://ftp.algonet.se/pub/index/yaz/>
[14] Nordic SR-Net: final report. Esbo: NORDINFO, 1994.
[15] Project information available at <URL:http://www.bibsys.no/one.html>
[16] Bernd Luchner. The DBV- OSI II project. Vine, no. 97, December 1994, pp. 37-38.
[17] Christine Bossmeyer and Bernd Luchner. DBV-OSI II: open communication between library and information retrieval systems. In: Information Superhighway: the role of librarians, information scientists and intermediaries. 17th Essen Symposium, 24-27 October 1994.
[18] John Favaro et al. ARCA user interface application requirements document. June 1995.
[19] Project information available at <URL:http://www.cranfield.ac.uk/eurilia/eurilia.htm>
[20] Rosemary Russell. Z39.50 and SR: an overview. Vine, no. 97, December 1994. pp. 3-8.
[21] Juliet Leeves with Rosemary Russell. Libsys.uk: a directory of library systems in the United Kingdom. London: LITC, 1995.
[22] The UK-ZIG mailing list is hosted by Mailbase. <URL:mailto:uk-zig@mailbase.ac.uk>. To join send a message with this text join uk-zig your-first-name your-last-name to mailbase@mailbase.ac.uk.
[23] The eLib information server is at <URL:http://ukoln.bath.ac.uk/elib/>.
[24] Ariadne <URL:http://ukoln.bath.ac.uk/ariadne/>.
[25] UK/European Z39.50 homepage <URL:http://ukoln.bath.ac.uk/Z3950/>
UKOLN (UK Office for Library & Information Networking), University of Bath, Bath BA2 7AY, UK. <URL:http://ukoln.bath.ac.uk/>