I found a note on USENET from someone at IBM asking if the
PKCS #7 OIDs were in fact wrong ... I checked, they sure seem
to be wrong. In class sun.security.pkcs.ContentInfo:

    private static int[] pkcs7 = {1, 2, 840, 1113549, 1, 7};
    private static int[] data = {1, 2, 840, 1113549, 1, 7, 1};
    private static int[] sdata = {1, 2, 840, 1113549, 1, 7, 2};
    private static int[] edata = {1, 2, 840, 1113549, 1, 7, 3};
    private static int[] sedata = {1, 2, 840, 1113549, 1, 7, 4};
    private static int[] ddata = {1, 2, 840, 1113549, 1, 7, 5};
    private static int[] crdata = {1, 2, 840, 1113549, 1, 7, 6};

Those "1113549" OIDs numbers should be "113549" instead (they have
an extra "1" up front). It's clear that basic interop testing for
this standard format didn't happen.