Steve and Isidro There are two points about links. The main point about links is that they are hardly used. Over 75% of entries to an average institutional repository comes from an out-of-repository search engine. The small number of uses of the in-repository search are by the local repository community. Entry by following links established by prior search or browse are more scattered, but in our case are easily dominated by accessing links from the University Library website. Next comes accesses by medical students from a med student website to our most popular document - a psychiatry textbook. I assert that none of these are critical to a repository's success.
I also note that my evidence points to people bookmarking a useful paper, rather than going to the trouble of writing it in a web-accessible page. The exceptions, as I noted before, are hyperlinks from the university's own bureaucracy, such as corporate staff pages, and the research website. For example all my papers are linked to from the research website. My publications are accessed via link to a browse by name facility. I suspect that only very few follow these links, which seems to be borne out by the evidence from the logs. Please note that I am not against using inward-bound links as one component of a rich set of metrics. It should be in there. But giving it a 50% weighting is absurd! 10% is more reasonable a priori. This brings me to the second point: Repositories were not set up to provide linkage, and if they were to be in the deep web apart from being harvestable, their utility would be only slightly weakened. Indeed this is exactly the situation with most of the PhDÂ thesis repositories in Australia. The federated site is open to the Web, and a very few thesis sites like my university's, but most university repositories are simply in the deep web, accessed only by the federated harvester. This is the Australasian Digital Theses Program, also listed in the Webometrics top 200. I haven't heard 30+ universities complaining about the loss of links. Arthur Sale Professor of Computer Science University of Tasmania > -----Original Message----- > From: American Scientist Open Access Forum > [mailto:AMERICAN-SCIENTIST-OPEN-ACCESS-FORUM@LISTSERVER.SIGMAX > I.ORG] On Behalf Of Steve Hitchcock > Sent: Tuesday, 12 February 2008 11:44 PM > To: american-scientist-open-access-fo...@listserver.sigmaxi.org > Subject: Re: [AMERICAN-SCIENTIST-OPEN-ACCESS-FORUM] New > Ranking of Central and Institutional Repositories > > I agree with most of Arthur's points, especially with regard > to activity and download measures, but I'm puzzled by his > comments about link-based visibility. He may be criticising > the method of calculation or its use in the overall > factoring, but in principle links seem a relevant measure for > repositories and one that should be factored in. >