The UK is leading the English-speaking world (hence the world) in research access today in a number of ways. The quadrennial Research Assessment Exercise (RAE) is not one of them, but if we're going to keep doing that too, we may as well do it right. So it's time to link our primacy in access to our preoccupation with assessment, for the benefit of both.
The link between access and assessment is productivity and impact. Scholars and scientists differ from other authors in that they publish their research findings (in refereed scholarly journals) not for royalties or fees, but for research "impact." They are rewarded (by their universities, research funding sources, prize committees and posterity) for their research productivity, which is assessed not on the basis of the sheer magnitude of their output (as "publish or perish" would suggest), but on the magnitude of its uptake: How much do their contributions affect the present and future course of research? Do other researchers use, cite and build upon them?
So making their work publicly accessible to all potential users has always been a necessary condition for everything else for researchers. That's what research publication is about. The only thing that tied this anomalous, giveaway world to the vastly different (and much bigger) non-givaway world of royalty/fee-based publication (books, magazines) in the Gutenberg era was the high cost of generating and disseminating print-on-paper. That real and inescapable cost had to be recovered somehow, and access-tolls were the only way to recover it, even though access-tolls meant access-barriers (to all potential users who could not afford to pay) -- which of course translated into impact-barriers, exactly the opposite of what research and researchers want and need.
In the PostGutenberg Galaxy, research in every discipline can and will continue to be refereed (peer-reviewed) as it always was, by a pyramid of established journals (currently at least 20,000) ranging from the most selective, highest-quality ones at the top, all the way down to nearly a vanity press at the bottom. It is the maintenance of this sign-posted hierarchy of quality that makes the research literature navigable and useable; and journal peer review is likewise the backbone of all assessment of research performance (for hiring, promotion, tenure, and other academic rewards). Hence it is also what the RAE, which determines universities' individual research funding levels, is based upon. If journals did not continue to implement peer review, not only would researchers not know which papers were worth reading and building upon, but universities would not know which researchers were worth hiring and promoting.
Yet peer reviewers perform their refereeing services for free, just as peer-reviewed authors report their research for free. This is all part of the much more indirect reward system of research, in which impact is the coin of the realm, not toll-gate income from text-sales. So journals' gate-keeping function (peer review), a service which controls and certifies research quality, should not be confused with their toll-gating function (subscription, site-license, pay-per-view S/L/P), which pays for an on-paper and/or on-line product, the text.
In the PostGutenberg Galaxy, the gate-keeping service remains essential, but what the toll-gating pays for (the on-paper/on-line text) has become a mere option. Yet the two are currently still wrapped together inextricably. The result is that the toll-gating for the inessentials (the publisher's version of the text) is blocking access to, hence the impact of, the refereed research itself.
Publishers cannot and will not remedy this on their own. Only authors and their institutions can eliminate all the obsolete and counterpreductive access/impact-barriers, virtually overnight, by self-archiving their own refereed research online, now. Free software with which universities can create their own interoperable "eprint" archives has been designed by Southampton University (http://www.eprints.org). These archives are all compliant with the Open Archives Initiative (http://www.openarchives.org), which means that all universities' distributed contents can be "harvested" into one global, virtual archive, free for all, worldwide, maximizing both the accessibility and the impact of all refereed research.
The RAE can help hasten the freeing of access to this literature by mandating that all UK universities self-archive all their annual refereed research in their own eprint archives. The harvesters (e.g., http://cite-base.ecs.soton.ac.uk/cgi-bin/search) will provide newer and richer measures of "impact" with the help of citation-linking services for open archives (http://opcit.eprints.org is an international collaboration between Southampton, Cornell and the Los Alamos National Laboratory, jointly supported by the NSF in the US and JISC in the UK). Not only the citation impact but the "hit" impact, for both the papers and the authors at UK universities and abroad, will be not only accessible but assessable continuously online by anyone who is interested, any time, instead of just in a quadrennial RAE exercise.