I agree with Simon on his point, > ... look for a solution that doesn't require creating a table every time someone thinks up a new label. , but I thought that was evident in the form of schema I suggested. However, I can see how that is not as obvious as I thought, leading to a need for that point. To be clear: I thought the requirement was to allow a slowly growing or changing (or "generally fixed" as I put it) set of labels to be associated with files. The labels table would contain that set.