Relies upon PIL to parse the initial JPEG structure and then does its own TIFF/Exif parsing because PIL's isn't up to dealing with the "full" Exif structure; looks like it gives up after IFD0.
Only tested with the V1 camera module, but I expect it'll work with either model. There's lots of stuff in here which isn't as efficient as it could be and a fair bit of cruft, but it's only a demo :)