the fallacy of using commercial identifiers in applications aimed at the general public
On the barcodepedia forum there's a discussion about using EANs in a database of 'ethical information' of products.
This is just one of number of efforts I've seen where people have tried to use a commercially developed identifier as the index into a system to be used by the general public, another being Jon Udell's LibraryLookup bookmarklet that let you search your local library's catalogue by ISBN.
But these repurposings always come up against the problem that the original identification scheme uses a much finer grained categorisation than the general public care about. So there is no single barcode for 'coca cola', there's thousands, varying not just by flavour (classic,cherry,zero,diet, diet with lemon, diet with lemon and a backrub) but also by country of manufacture, container size, whether the container is sold standalone or as a 6 pack, and lots of other variables that really matter to manufacturers, distributers, wholesalers and retailers of products (who created the UPC/EAN/GTIN scheme in the first place), but don't matter at all to someone wanting to discuss the ethics, nutritional value, or social history of coca cola.
Similarly, as Jon Udell discovered, an ISBN denotes a species, not a genus so there is no unique ISBN for "On The Road". And the need for members of the book trade to differentiate between books that have the same content but differ in publisher, edition or binding etc was one of the motivations for creating the ISBN system in the first place.
In the LibraryLookup case, the proposed solution is to use a service like xISBN takes an ISBN, and then returns a list of 'related' ISBNs, i.e. entries in an ISBN catalogue that have the same title.
And I suspect if anyone attempts to build a list of information about products for consumers and uses the GTIN barcode as the index, they will eventually hit the same problem of wanting to combine multiple GTINs into a single 'product', but the only way of combining multiple records will be by looking for items that have the same name.
So why not just index and search by name in the first place?
Clay Shirky wrote about this, and a whole lot more, (and a whole lot better) in "Ontology is Overrated"