If you remove the .doc extension from the filename of some application/msword documents, they are recognized as application/x-msi files. Attached are two testcases: good works as expected, bad does not. I've noticed that files edited with OpenOffice.org are correctly recognized.

Both files have the same start sequence, so the problem must be in other magic numbers... Word documents have a fairly rich set of patterns, which should be sufficient to detect them. You may notice that MSI files are a subclass of application/x-ole-storage too, that may be a hint.

I have not tested Excel and PowerPoint files, but they may suffer from the same issue. And BTW OpenXML documents are only detected a ZIP archives when without extension, but I'm not sure we can do anything about it...


For reference, the relevant parts of freedesktop.org.xml:

  <mime-type type="application/x-msi">
    <comment>Windows Installer package</comment>
    <sub-class-of type="application/x-ole-storage"/>
    <magic priority="50">
      <match value="\xD0\xCF\x11\xE0\xA1\xB1\x1A\xE1\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x3E\x00\x03\x00\xFE\xFF\x09\x00\x06" type="string" offset="0"/>
    </magic>
    <glob pattern="*.msi"/>
  </mime-type>

  <mime-type type="application/msword">
    <comment>Word document</comment>
    <sub-class-of type="application/x-ole-storage"/>
    <generic-icon name="x-office-document"/>
    <magic priority="50">
      <match value="\x31\xbe\x00\x00" type="string" offset="0"/>
      <match value="PO^Q`" type="string" offset="0"/>
      <match value="\376\067\0\043" type="string" offset="0"/>
      <match value="\333\245-\0\0\0" type="string" offset="0"/>
      <match value="MSWordDoc" type="string" offset="2112"/>
      <match value="MSWordDoc" type="string" offset="2108"/>
      <match value="Microsoft Word document data" type="string" offset="2112"/>
    </magic>
    <glob pattern="*.doc"/>
    <alias type="application/vnd.ms-word"/>
    <alias type="application/x-msword"/>
  </mime-type>