I run hot and cold on Dewey and the organization of knowledge. Over the weekend I wanted to find some books on web site security. The corresponding Library of Congress Subject Heading is "Web Sites — Security Measures". Looking this up in the catalog produced exactly eight books in the entire 40-library network, none at my local library, but one at a library 12.6 miles away as the google flies. When I got to the library I found two other useful books next to the book I was looking for, still in 005.8, but with slightly different subject classifications. I wouldn’t have found the other two books if I hadn’t been looking at an organized shelf.
In principle, I agree with David Weinberger. You should be able to search for and select books by arbitrary tags and labels from whatever source. If books had RFID, librarians could pile them in the middle of the library floor, or shelve them by size or color or recency, and patrons would still be able to find them. In practice, it’s too darn hard to browse using the current generation of discovery applications.
Weinberger is right that there is no right classification. The problem for libraries is that some organization is evidently useful. For librarians, you can put this in the form of two questions: what shelving order and what headings on the shelves? For patrons, the corresponding questions are: where is the book that I want, and what books are like this book? The Dewey shelf is alive because current software systems can’t replace how the shelf answers those questions.
Now astute readers might wonder why I didn’t search the catalog by Dewey number. It’s a good question. One reason is that it took me two days to think of it. Another reason is that the web interface to the catalog doesn’t really help the process. Now, click here, which is where you land when you click through to the catalog from the library web site. If you wanted to search by Dewey number, should you click on the "Advanced Search" link? Nope! It’s on the "Other" tab along with four other code number searches and search by journal title. (This last is an odd one, because it seems to be the same as an advanced search on title limited to journals.) But let’s do a search for items which match "005.8". Click here for the results. When you search by call number, the result is a list of matching call numbers. This isn’t helpful! No worst there is none.
There’s a genuine design problem here. Searches are often ambiguous, like a search for "web site security". A discovery application might choose to accept the ambiguity and display any matching items, or it might attempt to resolve the ambiguity. There’s no right answer. Amazon always shows you a list of items: if you search for "carol" on Amazon you’ll get songs by Carole King as well as books of carols in one unsorted list. On the other hand, if you search for "carol" on IMDB you’ll be shown carefully organized alternatives, including names and titles, as well as some politely phrased suggestions for "improving your search".
But there’s also an important principle which should drive the design of discovery applications: people using the catalog are looking for books (or CDs or DVDs or …), not catalog numbers or subject headings. The latter can only appear in the user interface if there’s a clear path to the former.
What the Dewey shelf provides that my library catalog doesn’t is a way to easily widen a search to related topics. You just keep looking along the shelf. You might miss other books shelved in other places, but you’ll see some related books. Weinberger might be right that the future is miscellaneous, but the present has been shelved in order.
Books and Libraries 