Some gotchas:

• The directory strucure is not obvious at all and it appears to have some variation from descriptions around the internet.

• The .epub file is just a zip file that has:

  1. A certain directory strucure (see below)

  2. A requirement that the first file, mimetype, to be stored uncompressed, have exactly the contents application/epub+zip with no trailing CR, LF, NL etc., and have no special attributes or extra fields.

  3. The tree hierarchy looks approximately like this:

      .
      ├── mimetype
      ├── META-INF
      │   └── container.xml
      └── OEBPS
          ├── book.ncx
          ├── book.opf
          └── chapter01.xhtml
     

• The file container.xml serves as a "bootstrap" for the book.

• The .opf file contains at a minimum, metadata of the book, the list of all files (except for mimetype, container.xml, and the .opf) that compose the book, with their full path, IDs and media-types, and, lastly, the "linear" list of topics in which the book should be read.

• The content of the dc:identifier element of the .opf file has to match the property of the meta element with name dtb:uid. Otherwise, http://www.threepress.org/document/epub-validate/ does not validate the document.