I take it as an assumption that increasingly books will be born digital, maintained digitally, and over time, due to enhanced content options and linkages, their preferred format will be digital.  Digital books will be available for (traditional ebook) device download; enhanced viewing and linking on networked screens (irrespective of glass size); print on demand; text to speech; and perhaps other services.   More and more publishers are beginning to populate their own repositories, and sell/rent space within their repositories to others.  Random, Harper, Macmillan, and Ingram (MyiLibrary) are all beginning to experiment with versions of these repository functions.  The better ones permit full text search across content.  

Even for publishers - because branding in publishing is fungible - search across content pools is probably a market-enhancing and revenue optimizing strategy.  It does not take too much hind-quarters resting-based cogitation to suspect that a publisher+search engine deal that would provide a discovery service across the great majority of texts, with user selection being driven back behind the walls represented by publishers' content repositories, is an appealing concept. 

In this environment, which for libraries reprises the transition to digital journals but at a later, more intensively networked stage in the evolution of online resources, this could spell Bad News in several ways.  Here are two.

Let's assume it is essentially game over for discovery, that user-based search happens primarily external to libraries, and that libraries are forced to license access to digital texts, with a possible option to buy at a stipend.  (I suspect this stipend would be high, because it would sunder the possibility of future revenues for the publishers or aggregators.)  This is arguably bad enough in itself for libraries, which have already given their libraries away to Google. How would libraries even be able to record that they have licensed this material?  What metadata would they obtain?  Would a traditional cataloging record even be useful?  As with journals, one might imagine that libraries have summary  metadata; OCLC has a complete record for the digital entity; and the library enters license terms into an electronic resource management system.  That's it.  So libraries hold print records in full detail, and stub records (with no full-text) of digital items that they probably do not own, and are paying to access.  If I were a user, I wouldn't see much reason to use library services to do discovery, although I still might visit the library to obtain materials.  

There is a flip side to licensed content, which is preservation.  Currently, preservation of digital journal content is, ur, uncertain.  Few libraries actually mount copies of journal content on local servers.  Many contracts have "perpetual access" clauses which are in varying degrees undefined, and largely untested; in the last 10 years we have not experienced a major organizational publishing failure with loss of content.  It will happen.  There is also a Mellon funded project called Portico, which is a sterling effort, but with incomplete coverage.  Finally, there is a distributed system called LOCKSS, which has signed up a certain number of publishers, but again is incomplete.  I should credit publishers, who recognize that libraries - silly they - put a value on safeguarding cultural and information assets belonging to the society at large.  Of course, libraries have to pay for that frivolity.  

Could these scenarios be replicated for books?  Perhaps.  LOCKSS could sign up publishers to safeguard books; a Portico v2 could archive digital monograph-based material; contracts could contain perpetual access clauses.  But would this work?  Digital books will consume additional resources for storage and maintenance, and the library community has not discussed at any length this additional burden.  I wonder if libraries have the scale - or could even support the scale in others - to generate a workable preservation system.  Portico works on a subscription model, for example, on both sides (publishers and libraries); free rider and sustainability problems remain largely unsolved.  Critically, subscriptions for preservation access,  combined with new licenses for access to digital books, will strap libraries far beyond their current liabilities.  Buying paper books, after all, is a one time cost with the addition of physical maintenance; this is a fraction of the likely life cost of maintaining digital licenses.  The curves could be purty durn scary.  

And then there are the standards issues.  I might assume - naively - that publishers would rally around a common format for digitally prepared and published books, such as the IDPF's standards, which have been embraced by actors such as Adobe's digitaleditions software.  However maintenance of digital standards is a shroud whipped away by the wind with its owner dancing wildly and frantically behind it.  The costs of format maintenance, conversions, and so forth - particularly as digital books are naturally enriched through linkages to other content and services - will be staggering if not contained or modestly defined.   

The movement of book content to publisher controlled digital repositories, with historical texts digitized by Google, and with discovery interfaces necessarily driven to search providers with significant scale, service linkages, and distributed indexing capabilities - the future of libraries in bookland could be rather bleak.   It would be appropriate, I think, for libraries to convene to discuss what the best approach might be to some of these scenarios - first among themselves only, and second approaching publishers directly, as both market-driven actors and historical partners.  After all, not even a failing publisher wants to be known, even as their death drives their collection to ashes, as a publisher who squandered the thoughts and discoveries of the men and women whose work they were privileged to make available to others, once upon a time.