There's a general problem that's usually overlooked in discussions of "the end of paper books" -- the role of publishers. They are not printers (though they may provide printing if they can do so more cheaply than farming it out) ... they are filters. They are supposed to "recognize the good" (often, but not always, as a simple matter of profitability) and "create the good" when an author is unfocused, ill-organized, or unskilled. They supply important expertise in the form of editors and publicists.
E-books promise to disintermediate both the publishers *and* the distributors, including current giants such as Amazon. All an author needs is a server, and not even that really if he/she is willing to let an aggregator get a small cut.
In doing so, the non-printing functions of the publishers are either minimized or ignored. This, in turn, likely means that some set of opinion-leaders will have to determine what is to be viewed as "the good" in some context or another. It may also mean that someone will want to acquire rights in order to "own" as oppose to merely "license" -- and these may well be the existing libraries, acting for the "public good."
It seems likely, in at least the short run, that they'll print copies for security's sake in case the author/aggregator's server goes down (or is simply forgotten). Whether they'll continue to do so past the next 50 years or so is a very open question, in my view. The issue will likely be one of preservation under the worst possible electronic disruption (i.e., a "Carrington Event").