Clay Shirky highlights a difference between a tagging approach and a metadata approach. Metadata on Amazon is contributed by publishers and Amazon themselves - and it seems that it is rather inconsistent, with some editions of books falling into the 'Gay', 'Erotica' types of classification whilst other editions of the same book don't. Customers can tag products with words of thei own choosing.

Amazon appears to have decided to remove some classifications, based on the metadata, from their sales ranking mechanism. An email they sent to Mark Probst suggested that this was a policy decision to protect their customer base. Classifications which have been affected include feminist literature, classics, LGBT literature and books on sex and disability. Amazon have apparently subsequently said that this is due to a glitch - a story people on Twitter are not believing.

So, Amazon has a problem - de-ranking products based on their classification system is causing annoyance (to put it