The results to date can be seen here. (Warning: the page on this link contains all the books, and so has over 2000 images. If it breaks your web browser, you get to keep the pieces.) That's having done all of the DVDs, graphic novels, computer games and large-format fiction, and the small-format fiction up to Ellison (plus random other unfiled books). Still got to do the rest of the small-format fiction and all of the non-fiction. Cover images are the correct one for the edition we've got, as far as is humanly possible, but are mostly not scans of the actual physical copy on the shelves. 1,984 books done so far, and I think probably 1,200-1,500 to go.
The cunning plan behind this is to make it easier to sell a thousand or so books on Amazon and make some more room on the shelves. But I've not yet tried out that part of Delicious Library's functionality.
For anyone doing the same, here are a few random notes and tips based on the experience so far.
- Barcode recognition is actually very good -- I'd say that 80%-90% scan immediately, and only 5% or so fail to scan after fiddling around a bit.
- UK barcodes always produce the right match, if they're found, which they generally are. US ones for books more than a couple of years old generally fail to match or more commonly match the wrong item -- one book thinks it's a pair of boxer shorts. This seems to be the case even if the default locale is set to the US, and I'm not sure of the reason for it. But it's not worth scanning the barcode on US books unless they're very new.
- Entering the ISBN is not much more effort than scanning the barcode, and is more reliable. It's also, obviously, the only option you've got for books published between around 1970 (when they started reliably using ISBNs) and 1990 (when barcodes came into common use).
- Always use the 10-digit ISBN, not the 13-digit one. Apart from saving three characters of typing, it's recognised more reliably.
- If there's a nine-digit number on an older book, try putting a 0 in front of it to make a ten-digit ISBN.
- It's very bad at matching up ISBNs with cover art -- for example, you'd think it could tell that a 1972 edition of Do Androids Dream of Electric Sheep is unlikely to have a Blade Runner movie tie-in cover.
- For non-recent books, Delicious Library is unlikely to find the cover art on Amazon. It's not worth using the built-in Google Image Search; go and search for the book on LibraryThing, which has a much wider range of cover art.
- A lot of paperbacks from the 1970s and earlier are starting to fall apart.
- It alphabetises authors on the basis that the last word before the end of the author's name (or before the first "&" for collaborations), is the author's surname (although it will ignore the odd suffix like "Jr."). This doesn't work for people like Ursula K Le Guin and Patrick Nielsen Hayden, and is irritating.
- Its book sizing algorithm is screwed. It has reasonable default sizes for "Mass Market Paperback", "Trade Paperback" and "Hardback", but most books are classified as "Paperback" which is sized rather larger than a normal paperback, and indeed larger than some trade paperbacks. (This doesn't show on the web version, where all books are the same size.)
- A lot of cover art gets truncated a bit at the top and bottom. It would be better if it changed the book's aspect ratio to match the art. (Again, not sure if this shows on the web version.)
- Always search the UK Amazon first, even for US books, as I suspect this will make it a lot easier to sell them on Amazon UK when the time comes.
- The book data has been entered inconsistently into Amazon by a lot of people, and needs some tidying up. Do it when you load the book, or you'll never do it. In particular, articles are missing from many book titles that start with them, authors names are occasionally reversed, and there's no consistency over whether it's Philip K Dick, Philip K. Dick or Philip K.Dick (which means you end up with three different authors).