I had similar woes. The way to identify encodings is not guaranteed unless a (knowledgable) human is employed. Corion proved that above. But check the comment there how firefox and chrome tackle this issue.