Linking is as easy as SFX

Jenny Walker explains how linking technology can overcome the frustrating problem of 'dead' or inappropriate links.

Linking is as pervasive as the web. Web users all recognise that clicking on hyperlinked text in a document will cause a link to be made to another document - or another section of the same document. Such links have revolutionised our world. When reading an online news article, we are now able to hyperlink to background reading, previous reporting on the subject, information about the news location, the current weather conditions there, and so on. Such links, which are intended to aid navigation and thereby enrich our experience, are embedded in the source document at the time of creation, either by the creator of the document, or by automated means.

Such ease of navigation through resources is now expected in a scholarly information environment. For example, users and librarians expect to link from an article citation to the full text of the article. With the explosive growth of web technologies, this expectation should be readily realised. However, the only linking solutions available until recently fall short for a number of reasons.

• Links are calculated and embedded in documents at the time of creation of the documents. To create these 'static' links, the information service provider must have knowledge, at the time, of all resources to be interlinked. This approach cannot easily be scaled up to deal with large numbers of resources, it may cause delays in the delivery of information and the resulting links may not be reliable.
• Solutions that rely on dynamic rather than static linking, whereby the links are calculated at the time the user requests the link, tend to be 'closed' solutions. Such solutions put control of the linking in the hands of the information service providers, which are focused on maximising the link traffic to their sites, rather than in the hands of the librarian, who would like to maximise usage of the varied resources for which the institution has already subscribed.
• The scope of the links offered to users is limited typically to full-text retrieval, or to holdings lookups in online catalogues. Little, if any, provision is made for a richer range of extended service links, such as those for citation database searches, related web searches and online bookstores.
• Links will appear in information resources regardless of whether users have rights of access to the information resource or service to which they are linking. Users may click on a link only to discover that they are barred from access to the service. This leads to frustration and in some cases may result in the user purchasing a service for which the institution has already paid, but via an alternate provider.

Fig. 1. In the 'old' way of linking, the resolution of links is determined by the link source.

The 'appropriate copy' problem
The last constraint is often referred to as the 'appropriate copy' problem. 1 Despite today's interlinked world where in theory it is possible to access a single version of a resource no matter where it is hosted in the world, it is still possible for many copies of an article to exist in electronic form. This is mostly a good thing as it provides for reductions in wide-area network traffic (which could incur costs) and for competition in business models. However, it might be that the appropriate copy of an article is not an electronic one for which the library has a subscription, but rather the print one in a journal on the library shelves. The appropriate copy could also be one that could be obtained through an ILL request to another institution, a request to an approved commercial document delivery supplier, or on a 'pay-per-view' service from the publisher or other provider. Further, not all electronic forms of documents available over the web are the same version, but may nevertheless be variously quoted as authoritative sources. Value judgments on 'appropriateness' are the preserve of the librarian.

Smart linking
For systems to offer the flexibility required so that librarians can ensure that links lead to the appropriate copy or indeed the appropriate service, a new way of thinking and a new way of linking - smart linking - is required.

In the late 1990s Herbert Van de Sompel and his team at Ghent University in Belgium developed a link server to address not only the 'appropriate copy' problem but also the other constraints on linking that were prevalent at the time, and largely remain today.2,3,4 This link server was called SFX - 'special effects' - as SFX linking would enliven our interaction with scholarly information. In February 2000, Ex Libris purchased the SFX link server technology from Ghent University (see www.sfxit.com/news/pr20000204.html ) and is now implementing SFX servers for libraries around the world.

Link servers, such as SFX, introduce a revolutionary step in the linking process by disconnecting the link source from the link target. The links provided in a source document no longer describe the target of the link - as is the case with embedded hyperlinks - but rather the source itself. In this way the source document provides a 'hook' that enables communication with a link server. Your library can then individually, via the link server, define the target(s) of the links based on your own requirements and local conditions. Thereby, your library can take control of its linking environment, either by supplementing or in some cases replacing the linking capabilities provided by the information providers.

Fig. 2. Smart linking is accomplished by disconnecting the link source from the link target, and inserting a link server, such as an SFX server, in the linking process. The link source provides a 'hook', and the link server accomplishes the context-sensitive resolution of links to their appropriate target(s).

Introducing new linking terminology
New terminology is required to describe new models for linking.

Link source: the information resource where the user begins, such as a citation database, online catalogue or journal article with references.

Link target: the information resource or service where the user lands as a result of having followed a link. This could be directly at the article on the target site, at a table of contents page, a journal home page; or it may be a set of results retrieved from a search on the link target.

Link services: options, decided by each library/link server installation, which are presented to the user after they click an SFX button to follow a link from a link source, such as a citation in an abstracting and indexing database. These may include:

• a link service to jump directly to the full text of the article on a publisher or aggregator site;
• a link service to search your library's Opac for the catalogue entry for the journal from which the article is cited, in order to check local holdings details and status;
• a link service to submit a request for the article from your library's preferred document delivery supplier;
• a link service to search for other articles by the author cited;
• a link service to search local information repositories;
• a link service to search relevant web-based resources, for example the RDN subject gateways;
• ....and many more.

Extended link services refer to those types of services that extend the notion of linking beyond linking just to the full text.5

Smart linking is accomplished by disconnecting the link source from the link target, inserting a link server, such as an SFX server, in the linking process. The link source provides a 'hook', and the link server accomplishes the context-sensitive resolution into links to the appropriate target(s) (Fig. 2).

The link server, being under the control of the local institution, can exert complete control over the types of link services that are offered to users at any time and to where these are resolved. In this new world of linking, the link source includes not a static or hard coded link, but rather a 'hook' which is inserted dynamically at the time the user views the link source and which provides information about the link source rather than the link target. A standard has already been defined for these linking 'hooks' - the OpenURL.

Fig. 3. User workflow for SFX smart linking - from an SFX link source via an SFX link services menu to an SFX link target, in this case the full text of the article.

The OpenURL
The OpenURL standard allows for descriptive metadata elements and identifiers to be transported from the link source to the link server. The OpenURL is implemented by the information provider in its service and can be applied by it at various levels within its resource. The OpenURL could for example, in an abstracting and indexing database, be offered to the user for each record shown in a brief results list and/or for each full record shown. When a user clicks on a button in the link source, an OpenURL is created and sent to the link server. An SFX button represents the OpenURL sent to an SFX server. The OpenURL has two key components:

1. The Base-URL that determines the address of the link server to which the OpenURL should be sent. Sites must provide this information to the information service providers for each link source. The notification can be automated.
2. The Query or 'content'. This can be descriptive metadata elements and/or identifiers.

The OpenURL, already a de facto standard for linking, is on a fast track path towards acceptance as a Niso standard (see http://library.caltech.edu/openurl). Pending acceptance by Niso, many information providers have already implemented the OpenURL according to the published draft standard, and many others have committed themselves to do so (see www.sfxit.com/sources.html).

The OpenURL is easy for information providers to implement, and it provides a cost-effective means of providing libraries with appropriate ways of linking across information resources irrespective of providers. Typically, information providers offer the OpenURL open linking capability alongside their default, resource-specific links.

An important illustration of the ongoing penetration of the OpenURL linking framework is the current prototype in which the OpenURL framework and concepts are integrated with the DOI/CrossRef linking solution. This prototype, in operation at Los Alamos, Ohio State University and the University of Illinois, demonstrates how the OpenURL linking framework can complement the DOI/CrossRef Framework by addressing the problem of the user's context, which is not taken into consideration in the delivery of CrossRef links, leading to immense user frustration. The current prototype has the support of important groups such as Pila (Publishers International Linking Association, Inc.), IDF (International DOI Foundation), DLF (US Digital Library Federation), CNRI (Corporation for National Research Initiatives).6, 7

* * *

Citation:
Mark P. Purdue, Loraine D. Marrett, Larry Peters, Jason K. Rivers Predictors of Sunburn among Canadian Adults. Preventive Medicine October 1, 2001 Vol. 33, No. 4, pp. 305-312 (doi:10.1006/pmed.2001.0888) Examples of possible OpenURLs that could be included by the information resource as a means to allow for open linking for the above citation.

http://demo.exlibrisgroup.com:8888/demo?aulast=Purdue&auinit=M&date=2001-10-01&stitle=Preventive Medicine&atitle=Predictors of Sunburn among Canadian Adults&volume=33&issue=4&spage=305

http://demo.exlibrisgroup.com:8888/demo?id=doi:10.1006/pmed.2001.0888

Legend:
bold - BASE-URL of linkserver

normal - metadata and identifiers

Table 1: Valid OpenURLs for a given citation

* * *

Fig. 4. An institutional SFX server permits linking among all parts of a library's electronic collection. The configuration of the SFX server determines the extent of the interlinked environment..

Configuring an SFX server
SFX, and other link servers, operate quite independently of the content itself, communicating directly with the systems providing the content.

The SFX services offered to users are only as good as the information provided by the librarian. The more detailed the information, the better the level of granularity for linking. If, for example, the librarian provides configuration details identifying the year or volume and issue from when a particular journal was made available in electronic form, then link services for this e-journal will only be shown to the user if they meet the necessary date or enumeration criteria.

User-friendly tools are provided to facilitate the task of configuring the link server. An SFX server comes with a comprehensive global KnowledgeBase that includes generic information about SFX targets such as how to link to a particular target, what journals are held on that target site and the extent of coverage. If the library subscribes to the base set of information, e.g. all journals from the Academic Press Ideal service, then it suffices simply to activate that target by toggling a Yes/No button. However, if the library subscribes to only some of the Academic Press journals or for only the current year - or indeed these journals are hosted locally - then such information will need to be reflected in the local SFX KnowledgeBase. If the information is held elsewhere by the library in electronic form - in a library catalogue or an Excel spreadsheet - then it should be possible to import this into the SFX local KnowledgeBase.

Documented evidence from some of the many SFX sites now installed indicates that the resource required by libraries to manage an SFX server on an ongoing basis is 0.25 FTE.8 However, this is unlikely to be an additional 0.25 FTE but rather an effective reallocation of staff resources already dedicated to the management of e-journal holdings and linking services.

Facilitating an article economy
New models for linking enable new types of services. With SFX, librarians can determine which resources to include in their interlinked environment and the manner in which they are interlinked. Librarians are free to use pre-defined SFX services or to create their own services, e.g. to link to a local data repository such as an electronic theses and dissertations database. As new services are offered by information providers, these can also be incorporated.

One emerging service links users to publisher and aggregator sites where the user can pay for services for which the library does not have a subscription. Libraries will take differing views on this approach, but there is no technological barrier.

'It's magic'
In the words of Judi Briden, Electronic Resources Co-ordinator at the University of Rochester in the US - 'It's magic!'9 However, some of this 'magic' - linking from a citation directly to the full text of the article cited - is what users have come to expect from their scholarly information systems. With SFX, users also experience the magic through serendipity. They may search for one particular item and are led, by association, to the discovery of other relevant and related material.

From the librarian's perspective the magic presents itself in different ways, depending on their role:

• For user services librarians: SFX allows for standardisation of link services across resources, reducing training needs. Users will see SFX buttons appearing in many, if not all, their resources. Clicking on an SFX button is intuitive.10
• For technical services librarians: SFX offers a single point of administration of the various services across different data sources. The SFX server is configured with 'knowledge' about all the resources to be interlinked and the SFX server then delivers up the links. No more uploading holdings files to many different service providers to enable linking.
• For library managers: SFX captures and reports a wealth of statistics on the use of distributed resources that has never before been available.

References

1 Priscilla Caplan and William Y. Arms. 'Reference linking for journal articles.' D-Lib Magazine. 5(7/8). Available at: www.dlib.org/dlib/july99/caplan/07caplan.html

2 Herbert Van de Sompel and Patrick Hochstenbach. 'Reference linking in a hybrid library environment. Part 1: frameworks for linking.' D-Lib Magazine, 5(4), 1999 (www.dlib.org/dlib/april99/van_de_sompel/04van_de_sompel-pt1.html).

3 Herbert Van de Sompel and Patrick Hochstenbach. 'Reference linking in a hybrid library environment. Part 2: SFX, a generic linking solution.' D-Lib Magazine, 5(4), 1999 (www.dlib.org/dlib/april99/van_de_sompel/04van_de_sompel-pt2.html).

4 Herbert Van de Sompel and Patrick Hochstenbach. 'Reference linking in a hybrid library environment. Part 3: Generalising the SFX solution in the "SFX@Ghent & SFX@LANL" experiment.' D-Lib Magazine, 5(10), 1999 (www.dlib.org/dlib/october99/van_de_sompel/10van_de_sompel.html).

5 See 2.

6 Herbert Van de Sompel and Oren Beit-Arie. 'Open linking in the scholarly information environment using the OpenURL framework.' D-Lib Magazine, 7(3),2001 (www.dlib.org/dlib/march01/vandesompel/03vandesompel.html).

7 Oren Beit-Arie et al, 2001. 'Linking to the appropriate copy: report of a DOI-based prototype'. D-Lib Magazine, 7(3), 2001 (www.dlib.org/dlib/september01/caplan/09caplan.html).

8 Miriam Blake. SFX - The Inside Issues for Libraries. Presentation by Miriam Blake of Los Alamos National Laboratory Research Library. LITA Electronic Publishing/Electronic Journals Interest Group Guided Discussion, ALA Midwinter, Washington DC, 14 January 2001. Available at: (www.sfxit.com/publications/presentations/SFXALAJan2001Mblake.ppt)

9 Judi Briden. Implementing SFX @ University of Rochester. Presentation by Judi Briden, NERCOMP conference, September 2001. Available at: www.sfxit.com/publications/presentations/SFXNRCMP_share.ppt

10 Miriam Blake. SFX at Los Alamos. Presentation by Miriam Blake at ALA June 2001. Available at: www.sfxit.com/publications/presentations/SFXALAjun2001Mblake.zip

11 Arthur C. Clarke. Profiles of the Future. 1962.

Jenny Walker is Director of Sales & Marketing, Information Services Division, Ex Libris (USA) Inc, 100 Galen Street, Watertown, MA 02460 (Tel: +1 617 923 0160 x526; email: jwalker@exlibris-usa.com).

© The Library Association
Last updated: 14 November 2001