*** Bug 759 has been marked as a duplicate of this bug. ***
This has not been fixed at all. We still work with a pointer to tags.
If we have 256 different tags for all songs then we need 32 bytes per song to store this information if we would use a bit pattern. At the moment we store each tag as a pointer to tags (which are of type unsigned2). So if we have 1.5 tags per song we get 11 bytes per song. If we have 20000 songs this change would result in 410kB extra memory. This memory must be accessed when scanning for songs so the question is whether this would indeed go faster. Probably not, since we have now very few checks for each song.