In my original document I identified the alternative to “open access” journals with page charges as the arXiv and institutional repositories, simply because this was the alternative considered in the EPSRC announcement. But as Laci points out, there are many journals which are either free (e.g. Electronic Journal of Combinatorics) or comparatively low-cost (e.g. journals from learned societies such as the London Mathematical Society). The first option is ideal in many respects but puts heavy demands on the volunteers who run the journals (and deserve the gratitude of the community for this). The second solution is also important. But there is an issue here too. At present the learned societies derive much of their income from their publications, especially journals. They may be vulnerable to a change in the model for funding journals or to pressures to make them freely available.