In the mid term (3 to 5 years), the electronic subdivision should target schools, universities, corporations, states, and other large organizations to accept subsets of their electronic catalog at small cost per book, but at large cost per catalog. These organizations will serve as suitable testing grounds for the new techniques on a large scale. This stage will also make for word-of-mouth advertising.
Large organizations will not allow wholesale copying for fear of lawsuits and public embarrassment. It is easy to ensure that by inserting hard to find identifying tags in the particular version of the books they receive; so if excessive copying is suspected there is a way to prove that a particular copy was made from the files the publisher supplied to the organization. To be able to insert these tags the subdivision needs to develop, or at least have a say in, the software used to display books; and that will happen only if the publisher is one of the first of the new publishers. This phase should generate a fair amount of capital that can then be plowed into development for the final phase.
At this stage publishers can work out ways to divide the income; the royalty system may have to change. With so many dollars coming in every year (or quarter, month, or day, depending on which is the better business policy), non-fiction publishers can afford to pay authors an advance the same way fiction houses do now, then keep track of demand for their work crediting them for any accesses of their work by any subscriber. Or publishers can buy titles on consignment. Or capital-heavy publishers can buy a work outright.
Other arrangements are possible, for example, capital-light publishers can ask authors to pay for the privilege of being put on their list, as vanity presses do today. In this scheme, authors bet that the demand for their work, once it is widely available, will outweigh their capital outlay; a better system than the present one if authors have the capital. Since the publisher's marginal distribution cost is near zero the publisher risks little. To avoid simple frauds later when an author can also be a subscriber, each subscriber's electronic connection can be recorded and publishers can require them to identify themselves to gain access.