Uploading files from web app, storing to temp location, extracting metadata/text prior to client approval/edit and committal to document repository.
MagicMimeTypeIdentifier only works for Open Office files if they retain their original extensions (.odt, .ods etc). Otherwise it detects them as a zip file (application/zip) therefore the extractors don't work. For PDF's and MS Office documents on the other hand it works fine, extensions irrelevant.
In my case get the extension from the file upload and use that for temporary file names, but that obviously won't work for identifying arbitrary binaries.