According to ISBN official standards, each format of an e-book should be given its own ISBN. This means if a book is sold in mobi-pocket and Adobe formats each would be given a separate (unique) number by the publisher even if the content is exactly the same. During the revision process for the current standard, this point received intense discussion mostly focused on the burden that applying what could amount to several hundred ISBNs to a single work would have on publishers' processes. We resolved this issue for the standard with judicious use of words such as 'shall' and 'should' but the issue was raised again recently when the ISBN board released a 'policy statement' reaffirming the need for separate ISBN's on each format of an eb0ok.
The reasons for this action is simple. Downstream supply chain business such as wholesalers, distributors and retailers require a unique reference to all products that pass through their operations. If one doesn't exist these businesses tend to apply their own numbers. In actuality, the practice of downstream partners applying their own numbers has been going on since the establishment of ISBN and isn't unique to e-books, but the issue is coalescing now around the obligations of a publisher to 'correctly apply' the ISBN standard to e-books.
At a meeting this week at AAP NYC a number of publishers expressed doubts about the need for this requirement. As a participant in the revision of the standard my view was simple. A publisher should want to manage and control the meta-data associated with all their products and enabling - by omission - the need for someone else to apply their own information never seemed prudent to me. Secondly, the veracity of the ISBN system is brought into question if more than one entity applies separate numbers to the same content. This occurs if B&N and Amazon sell the same e-book in the same format but in the absence of a publisher number they apply their own identifier.
At least one major publisher at the AAP meeting is not following the standard and after several years of distributing e-books and applying one ISBN irrespective of format (.epub for example) they are seeing no issues with confusion or misuse of their meta data. This is a powerful argument and comes from a publisher that is highly protective of their bibliographic information. If reflective of a general consensus the ISBN board should reconsider the wording of there directive. For example, simply changing the wording by inserting the words 'publishers may apply ISBNs to separate formats' would give enough latitude to those publishers that see a need to apply separate ISBNs and those that do not.
There are several qualifications (and others may raise more). Firstly, the issue of downstream partners which need identifiers for their internal process requirements must be governed. For example, in those cases where a publisher expects detailed sell-thru data they may provide ISBN's. If a downstream partner can only use a 13 digit identifier in their systems the publisher may require the partner to use an ISBN provided by the publisher. If the partner can use a non-ISBN (but NOT a dummy ISBN/13 digit id) such as letters and numbers the publisher may see no need to apply ISBN's. Secondly, the danger that rogue ISBNs that are intended to operate only within the operating systems of specific partners (wholesalers, vendors, etc.) escape into the supply chain causing confusion and much remediation is a real one and should be recognised. Currently, there aren't that many e-books and there aren't that many publishers working outside the recommendations of the standard. As e-books explode in distribution, data integrity problems that are virtually non-existent today may become very relevant issues very quickly.
Lastly, in a supply chain world where suppliers and retailers are racing (admittedly not a sprint more a marathon) to apply unique identifiers on individual items via RFID, this discussion runs counter to the logic other more sophisticated industries are following. Quite rightly, with volumes as small as they are, it may not be interesting to know which e-book versions seem to perform better, or get less customer service/help desk calls, or which package of products seem to show up on what platform or which segment of buyers seems to have what behavioral characteristics, or which partner seems to sell what types of products or formats, or which formats tend to be pirated more or less, and on and on and on. As the chain becomes flatter - as it is - publishers are going to want to know this stuff and tying a user to a format may be critical to all aspects of what they do.