The case for distributed over central OA archiving

Stevan Harnad, Central versus Distributed Archives, Open Access Archivangelism, December 16, 2006.  Excerpt:

Arxiv has been showing this same, steady, unswerving linear increase in the number of deposits per month (quadratic acceleration of the total content) since the year 1991, and Arxiv has been tracking its own growth, monthly, since then....

At this rate, Ebs Hilf estimates that it would take till 2050 to reach 100% OA in Physics. And that is without mentioning that Arxiv-style central self-archiving has not yet caught on in any other field (except possibly economics) since 1991. In contrast, distributed self-archiving in, for example, computer science, has already long overtaken Arxiv-style central self-archiving. See Citeseer (a harvester of locally self-archived papers in computer science, already twice the size of Arxiv):

Logic alone should alert us that ever since Institutional IRs and Central CRs became completely equivalent and interoperable, and seamlessly harvestable and integrable, with the OAI protocol of 1999, the days of CRs were numbered.

It makes no sense for institutional researchers either to deposit only in a CR instead of their own IR, or to double-deposit (in their own IR plus CRs, such as PubMed Central). The direct deposits will be in the natural locus, the researcher's own IR. And then CRs will harvest, as Citeseer, OAister -- and, for that matter, Google and Google Scholar -- do.

OA self-archiving is meant to be done in the interests of the impact, visibility, and recording of each institution's research output. Institutional self-archiving tiles all of OA space (whereas CRs would have to criss-cross all disciplines, willy-nilly, redundantly, and arbitrarily).

Most important, institutions, being the primary research providers, have the most direct stake in maximising -- and the most direct means of monitoring -- the self-archiving of their own research output. Hence institutional self-archiving mandates -- reinforced by research funder self-archiving mandates -- will see to it that institutional research output is deposited in its natural, optimal locus: each institution's own IR (twinned and mirrored for redundancy and preservation). CRs (subject-based, multi-subject, national, or any other combination that might be judged useful) can then harvest from the distributed network of IRs.