The following information describes the download path index files that are
used by PCBoard.  Lines marked with a vertical bar (|) were added on 7/5/96
in support of PCBoard v15.3's new style IDX file.

| NOTE:  Beginning with PCBoard v15.3, two IDX formats are supported by
|        PCBoard.  Both formats are "similar", enough so that you can share
|        most of the code in handling them.  The new format is mostly
|        different in that it uses 4-byte long integers instead of the 2-byte
|        short integers used in the old format. It also includes in the file
|        record portion of the file a 4-byte long integer representing the
|        size of the file.
|
|        The new style of index is created by running MAKEIDX with a /E
|        command line parameter.  Most sysops will probably continue to use
|        the old style index while reserving the new style index primarily
|        for CD-ROMs.


  File Specification
  ==================

  Old Style IDX Header
  --------------------
   2 bytes for an integer indicating the number of files
  52 bytes for 26 integers indicating starting points for A thru Z
| 73 bytes reserved space
|  1 byte as a "ID" value which is 0 to indicate the old style index

  typedef struct {
    unsigned NumFiles;
    unsigned RecOffset[26];
    char     Reserved[73];        // set to all zeroes
    char     Id;                  // set to 0 - indicates old style index
  } idxhdrtype;

| New Style IDX Header
| --------------------
|   4 bytes for a long integer indicating the number of files
| 104 bytes for 26 long integers indicating starting points for A thru Z
|  19 bytes reserved space
|  1 byte as a "ID" value which is 1 to indicate the NEW style index
|
| typedef struct {
|   unsigned long NumFiles;
|   unsigned long RecOffset[26];
|   char          Reserved[19];   // set to all zeroes
|   char          Id;             // set to 1 - indicates NEW style index
| } newidxhdrtype;


  Old Style File Name Records
  ---------------------------
  The filename records include:
  -  11 bytes for a filename of the form:  nnnnnnnneee
     any character not used is filled with a space
  -  the path associated with the file is represented by a path number

  typedef struct {
    char     Name[8];
    char     Ext[3];
    unsigned PathNum;
  } nametype;


| New Style File Name Records
| ---------------------------
| The new style record is the same as above with two differences:
| -  The PathNum variable is a 4-byte long integer
| -  In the new style index, a 4-byte long integer is included to indicate
|    the size of the file
|
| typedef struct {
|   char          Name[8];
|   char          Ext[3];
|   unsigned long PathNum;
|   unsigned long FileSize;
| } newnametype;


  Path Records (the same in both IDX styles)
  ==========================================
  Paths are stored as fixed length records, each one is 64 bytes which includes
  a NUL terminator at the end of the path:

  typedef struct {
    char Path[64];
  } pathtype;


  File Layout and Accessing the File
  ==================================
| The first thing you need to do is read the header of the IDX file and look
| at the byte at offset 127, which is the ID byte.  If it is 0, then use the
| structure defined as idxhdrtype shown above.  If it is a 1, then use the
| structure defined as newidxhdrtype.  The remainder of the "logic" for
| accessing the files remains the same whether using the old or new style IDX
| file.  Just remember to use the proper integer sizes (short in old style and
| long in the new style).

  Following the header is the list of filenames which are then followed by
  paths.

  The header consists of 26 separate values which indicate the record number
  (offset 0) of the first filename starting with each of the letters A thru Z.
  This reduces the number of file seeks performed by immediately narrowing down
  the search records.  It also increases the likelyhood that all file seeks
  will be performed within the memory buffer that is read in from disk.

  For instance, if the file to be found starts with the letter "B" then looking
  at offset numbers for "B" and "C" (let's say they are record numbers 100 and
  200 respectively out of 1000 files) we can narrow down the search to records
  100-200 immediately without having to read files at the very end of the list.

  Files starting with letters before 'A' will be placed immediately following
  the header so no offset value is recorded there.  Files that start with
  letters after 'Z' will follow those filenames starting with 'Z'.

  A binary search is then used to locate the filename.  Note the format of
  the filenames which makes it easy to use a wildcard matching technique.