My line of work is research. Research, for the most part, is not something you can sell on the private market. Therefore, one can't evaluate the quality of a professor's research using sales, profits, or anything like that. Everything is based on reputation -- what do important people think about the quality of your work?
Traditionally, the list of "important people" started with editors -- editors serving academic book publishers or journals. The editors, in turn, rely on the judgment of other people they trust -- reviewers -- to help them decide whether a submission was good enough to publish. And publication was the key to getting your work on the desks of other important people -- subscribers. The journals and presses who provided this review service managed to break even, or in some cases turn tidy profits, by exploiting their monopoly power over content.
Today, the review process is no longer the gateway to dissemination. And this, in turn, threatens the viability of the entire system. The widespread availability of journal articles -- for example, on researchers' web pages -- dilutes monopoly power. The ease of electronic distribution has encouraged new startups -- PLoS in the sciences, for example -- who offer content at no charge. Some academic disciplines have de-emphasized journal publications. This further endangers the financial standing of traditional publishers.
Not only is the review process unnecessary for dissemination, it is also not entirely necessary to establish the quality of work. Instead of inferring the quality of work by whether some journal editor liked it, we can count up the number of times it has been downloaded, or the number of times it has been cited in subsequent work. The wisdom of crowds replaces the sometimes idiosyncratic tastes of one person. All in close to real time. The economic necessity of journals & presses, in sum, has been mostly eliminated by advances in technology.
There's still one service that the journals can provide that the web can't. They can help authors to improve the quality of their work. They can point out broader implications that the author might not have recognized. They can help improve exposition, so that readers understand points more completely. And they can suggest directions for further work.
These services are of clear value to society. Who should pay for them? The standard economic answer is "whoever receives the benefits." Some of them accrue to society at large; but the clearest beneficiary is the author, who earns a better reputation through the process. So this suggests that the future of academic publishing involves shifting costs from subscribers to authors.
The future is already here. The PLoS model -- which has enabled the organization to break even -- involves charging nothing to subscribers, but imposing fees on authors of published work (to the tune of nearly $3,000 per article, albeit with waivers for those without the ability to pay). This might not be perfectly optimal -- some of the benefits associated with the review process accrue to authors whose work is ultimately not published by the journal. But it is probably closer to the optimum than the traditional model of charging little or nothing to authors, but outrageous subscription fees.