Sir

Referencing websites for supplementary data and online applications is increasingly popular and acceptable in today's scientific publishing. Yet my recent experience of testing the validity of web links referenced in papers published in 1998 supports the view that many websites are valid for only a short time.

I used PubMed to search for papers using the string “http”, and found about 300 citations. Of the 265 web links that appeared in abstracts, only 150 (57%) are still valid; 97 (36%) are no longer available and 18 (7%) redirected me to other URLs. Although these statistics are biased towards biomedical research and might not reflect other disciplines, they are discouraging.

The root of the problem is the formidable challenge of maintaining a website indefinitely. Many websites are stored on local servers, managed by system and network administrators who are not necessarily aware of the need to maintain links cited in publications in unaltered form. Other problems abound: for example, individual researchers might lose data in computer breakdowns, and the ownership of Internet domain names is often out of their control.

Authors must therefore take great care when citing web links. They must ensure that any website they create to supply additional information will be maintained indefinitely before referencing it in their publications. If authors need to reference websites belonging to other organizations or individuals, they should try to use only those that have been cited in other publications, as these sites are usually better maintained. Authors should redirect outdated web links to current URLs. Publishers should encourage authors to deposit their supplementary information in databases that are maintained carefully by the publishers. Many publishers now use the Digital Object Identifier (DOI) system to guarantee linkage, but this does not solve the problem of links to supplementary information in various academic or other informal websites. Further, DOI is relatively expensive for academics to set up on their own websites.

Given the convenience of the Internet, more and more information will be stored digitally as time goes on and URL references will become increasingly common. The lifetime of those references will determine how often we face a missing link.