/* File structure: (assuming the id-string is "SDF format"):

 * 1.  Bytes 0-10 - file identifier string, including null char terminator 
 * 2.  Bytes 11-18 - 64-bit integer, containing file position of next byte
 *     in the header.  Byte 11 is also strlen(sdfstr)+1 .
 * 3.  Bytes 19-26 - 64-bit integer, containing file position of next byte
       in the data portion of the file (which starts at byte 2019).
       Byte 19 is also strlen(sdfstr)+1+sizeof(pos) .  The value of this
       integer should also be the filesize.
 * 4.  Bytes 27-30 - 32-bit integer, containing the number of datasets 
       currently written in the file.  Byte 27 is also
       strlen(sdfstr)+1 +2*sizeof(pos) .
 * 5.  Bytes 31-38 - 64-bit integer, containing the size of the file header,
       'hdrsize', counting from byte 19.  Initially this is set to 2000, 
        but is increased in blocks of 2000 bytes as needed.
 * 6.  Bytes 39-2018 - A series of linefeed terminated strings, containing
       descriptions of the data to be written.  These strings generated by
       the data_to_string function.  Byte 39 is also
       strlen(sdfstr)+1 + 2*sizeof(pos) + sizeof(i4) *sizeof(pos) .
 * 7.  Bytes 2019-eof - the data, with each dataset written squentially,
       in order.  Byte 2019 is also equal to
       strlen(sdfstr)+1+sizeof(pos)+hdrsize .  If hdrsize is larger than
       2000, the data will start correspondingly later in the file
       
       The integers in the header and the data are all written in large-endian
       byte order, if the data isn't simply byte data. */