README.DOS

Some notes about the supplied MSDOS executables and their compilers:

A) The 32-bit DOS executables "zip.exe" and the auxilary utilities
  "zipnote.exe", "zipsplit.exe", "zipcloak.exe" (crypt-enabled distribution
  only) were compiled with DJGPP v 2.03, using gcc 3.4.3 as compiler.
  They require a DPMI server to run, e.g.: a DOS command prompt window
  under WINDOWS 3.x or Windows 9x.  To use this program under plain DOS,
  you should install the free (GPL'ed) DPMI server CWSDPMI.EXE.  Look
  for "csdpmi5b.zip" under "simtelnet/gnu/djgpp/v2misc/" on the SimTelNet
  home site "ftp.cdrom.com" or any mirror site of the SimtelNet archive.

  We have decided to provide 32-bit rather than 16-bit executables of
  the auxilary tools for the following reasons:
  - Nowadays, it has become quite unlikely to find PC computers "in action"
    that do not at least have an i386 CPU.
  - the 32-bit versions do not impose additional archive handling limits
    beyond those defined by the Zip archive format
  - the 32-bit DJGPP executables can handle long filenames on systems running
    Windows 95/98/Me and Windows 2K/XP/2K3.

B) There are two 16-bit MSDOS executables provided in zip2?x.zip:
    zip16.exe     regular Zip program, requires ca. 455 KBytes of contiguous
                  free DOS memory or more.
    zip16-sm.exe  a variant that was compiled with the SMALL_MEM option
                  for minimal memory consumption; requires at minimum
                  322 KBytes of contiguous free DOS memory.

  The SMALL_MEM variant requires much less space for the compression
  buffers, but at the cost of some compression efficiency.
  Therefore, we recommend to use the "SMALL_MEM" 16-bit "zip16-sm.exe" only
  in case of "out of memory" problems (DOS memory is low and/or very large
  number of archive entries), when the 32-bit program cannot be used for
  some reason (286 or older; no DPMI server; ...).

C) Hard limits of the Zip archive format:
  Number of entries in Zip archive:            64 k (2^16 - 1 entries)
  Compressed size of archive entry:            4 GByte (2^32 - 1 Bytes)
  Uncompressed size of entry:                  4 GByte (2^32 - 1 Bytes)
  Size of single-volume Zip archive:           4 GByte (2^32 - 1 Bytes)
  Per-volume size of multi-volume archives:    4 GByte (2^32 - 1 Bytes)
  Number of parts for multi-volume archives:   64 k (1^16 - 1 parts)
  Total size of multi-volume archive:          256 TByte (4G * 64k)

  The number of archive entries and of multivolume parts are limited by
  the structure of the "end-of-central-directory" record, where the these
  numbers are stored in 2-Byte fields.

  Some Zip and/or UnZip implementations (for example Info-ZIP's) allow
  handling of archives with more than 64k entries.  (The information
  from "number of entries" field in the "end-of-central-directory" record
  is not really neccessary to retrieve the contents of a Zip archive;
  it should rather be used for consistency checks.) Also note that the
  latest Info-ZIP versions (currently in beta) now support Zip64 which
  extends all these limits, though these versions require file system
  support of large files.  MSDOS stores file sizes in 32-bit numbers
  internally. So, for MSDOS, the Zip64 features extending file sizes
  beyond the 32-bit barrier are not applicable.  It is possible to
  support Zip64 archives (for listing/testing, but not fully for
  creation/extraction to disk) on MSDOS using splits, but currently
  we are not planning on doing it.  The Zip64 extensions concerning
  "more than 64k archive entries" may be supported on MSDOS (at least
  up to the limit of 2^32 - 1 entries), unless we decide to drop support
  for the MSDOS port completely.

  Length of an archive entry name:             64 kByte (2^16 - 1)
  Length of archive member comment:            64 kByte (2^16 - 1)
  Total length of "extra field":               64 kByte (2^16 - 1)
  Length of a single e.f. block:               64 kByte (2^16 - 1)
  Length of archive comment:                   64 KByte (2^16 - 1)

  Additional limitation claimed by PKWARE:
    Size of local-header structure (fixed fields of 30 Bytes + filename
     local extra field):                     < 64 kByte
    Size of central-directory structure (46 Bytes + filename +
     central extra field + member comment):  < 64 kByte

D) Implementation limits of the Zip executables:

1. Size limits caused by file I/O and compression handling:
  Size of Zip archive:                 2 GByte (2^31 - 1 Bytes)
  Compressed size of archive entry:    2 GByte (2^31 - 1 Bytes)
  Uncompressed size of entry:          2 GByte (2^31 - 1 Bytes),
                                       (could/should be 4 GBytes...)
  Multi-volume archive creation is not supported.

2. Limits caused by handling of archive contents lists

2.1. Number of archive entries (freshen, update, delete)
    a) 16-bit executable:              64k (2^16 -1) or 32k (2^15 - 1),
                                       (unsigned vs. signed type of size_t)
    a1) 16-bit executable:             <16k ((2^16)/4)
        (The smaller limit a1) results from the array size limit of
        the "qsort()" function.)
        32-bit executables             <1G ((2^32)/4)
        (usual system limit of the "qsort()" function on 32-bit systems)

    b) stack space needed by qsort to sort list of archive entries

    NOTE: In the current executables, overflows of limits a) and b) are NOT
          checked!

    c) amount of free memory to hold "central directory information" of
       all archive entries; one entry needs:
       96 bytes (32-bit) resp. 80 bytes (16-bit)
       + 3 * length of entry name
       + length of zip entry comment (when present)
       + length of extra field(s) (when present, e.g.: UT needs 9 bytes)
       + some bytes for book-keeping of memory allocation

  Conclusion:
    For systems with limited memory space (MSDOS, small AMIGAs, other
    environments without virtual memory), the number of archive entries
    is most often limited by condition c).  For example, with approx.
    100 kBytes of free memory after loading and initializing the program,
    a 16-bit DOS Zip cannot process more than 600 to 1000 (+) archive
    entries.  (For the 16-bit Windows DLL or the 16-bit OS/2 port, limit
    c) is less important because Windows or OS/2 executables are not
    restricted to the 1024k area of real mode memory.  These 16-bit ports
    are limited by conditions a1) and b), say: at maximum approx. 16000
    entries!)

    Note that the Win32 versions in general support greater limits than
    the MSDOS versions.  If you are using zip from the command line in
    Windows then you are usually better off using the Win32 versions.
    Also, we are currently working on betas for UnZip 6.00 and Zip 3.0
    that support Zip64 which extends these limits. However, the Zip64
    features (with exception of the "more than 64k archive entries"
    extension) may never be supported on MSDOS (or any DOS emulation)
    because of its hard file size limit at 4 GByte.  It is possible
    to support some of the Zip64 features in split archives but we are
    currently not planning to do it.


2.2. Number of "new" entries (add operation)
    In addition to the restrictions above (2.1.), the following limits
    caused by the handling of the "new files" list apply:

    a) 16-bit executable:              <16k ((2^64)/4)

    b) stack size required for "qsort" operation on "new entries" list.

    NOTE: In the current executables, the overflow checks for these limits
          are missing!

    c) amount of free memory to hold the directory info list for new entries;
       one entry needs:
       24 bytes (32-bit) resp. 22 bytes (16-bit)
       + 3 * length of filename


Please report any problems at:  www.info-zip.org

Last updated:  19 February 2005