This directory contains executables for Tandem NSK.

  5640 Feb 22 2004  README          what you're reading right now
332563 Feb 23 2002  unz550xK.zip    UnZip 5.5, Tandem exes/docs, zipfile format
331671 Jan  9 2003  unz551xK-beta.zip UnZip 5.51a Tandem exes/docs   [BUGFIX]
507904 Feb 22 2002  unzip550.bare   UnZip 5.5, bare Tandem executable
194679 Dec 25 1999  zip23xK.zip     Zip 2.3, Tandem exes/docs (no encryption)

Encryption binaries are only available from our European site,
ftp://ftp.icce.rug.nl/infozip/TANDEM/ :

  [none available]

All Info-ZIP packages contain documentation.  The sources are in ../src .

The bare UnZip executable can be used to unpack the UnZip and Zip distribution
archives.  The UnZip archive contains two binaries, UNZIP and UNZIPU; the
former requires SUPER.SUPER privileges to run and can reset last-open/modified
timestamps, whereas the latter can be run without any special privileges.
Both were compiled using Tandem NonStop D48.  Thanks to Dave D. Smith
(smithdt@bp.com) for doing the Tandem ports and providing the executables.

IMPORTANT NOTE:  The official 5.50 executables have a bug in their support
    of text files; specifically, they write only the last block to disk
    (modulo 64 KB).  Binary files are unaffected, and all of the text files
    in the UnZip archives are less than 64 KB in size, so the existing 5.50
    executables are sufficient to unpack the 5.51a beta.  Note that this is
    a temporary bug-fix release; official 5.51 binaries will be available
    after 5.51 is released.


Some older notes from Dave:

    Note that this version of UnZip can only create unstructured, object
    and text files, not Enscribe files (Tandem's hierarchical database
    system, allowing key-sequenced files).  I often use the PC PKZIP
    program to zip things up and then use the Tandem to unzip - or vice
    versa - with no problems (other than the file-type limitation).  Each
    Tandem file has a file code that indicates (roughly) what the file
    contains:

          0     Enscribe or Unstructured
        100     Object
        101     Edit (ASCII)
        180     'C' Text file

[...]

    The Tandem does not support subdirectories.  It has a fixed structure of

        \<system>.$<disk>.<subvol>.<file>

    where all the names have a maximum of 8 characters.  Note that you can
    default the system disk and subvol parts to blank.  Thus all the
    following are valid file names:

        \NODE1.$DATA8.DB.FILE1
        $DATA8.DB.FILE1
        DB.FILE1
        FILE1
        $DATA8.FILE1            (this form is not recommended)

    I've mapped this structure into the zipfile format as:

        //<node>/<disk>/<subvol>/<file/

    Thus the above files would appear as

        //node1/data8/db/file1
        /data8/db/file1
        /db/file1
        file1
        /data8/subvol/file1     (subvol part picked up from current defaults)

    In addition since Tandem does not support file extensions I have added
    the ability to specify extension names when zipping and storing as part
    of the ZIP internal name.  For example:

        ZIP zipfile "zipdir.readme txt"

    is stored in the zipfile as /zipdir/readme.txt

        ZIP zipfile zipdir.readme

    as stored as /zipdir/readme

    The opposite operation is performed for UnZip - assuming that the
    mapping operation is OK.  In the event that this fails it just uses
    the filename part of the name and throws away the directory information.

[...]

    On Tandem (NSK) systems there are two sorts of processors - CISC and
    RISC.  As you might expect, all current machines use RISC processors.
    However, until very recently all compilers produced CISC code.   This
    CISC code runs on both CISC and RISC machines.  When running CISC code
    on RISC machines it gets interpreted - which obviously slows things
    down.

    To get around this Tandem introduced a product called AXCEL that
    'accelerates' the CISC code.  It adds an extra RISC code region to
    the object which makes the object a lot larger (factor of 3) but runs
    a lot faster.  Programs such as Zip/UnZip that perform a lot of data
    manipulation run exponentially faster after being AXCELed.

    In addition on Tandem the C compiler has a number of selectable memory
    models - currently SMALL, LARGE and WIDE.  I have chosen the LARGE
    memory model as it means the object runs slightly faster.  Selecting
    WIDE allows larger buffer sizes, as 'int' becomes 32-bit rather than
    16, but still seemed slower when I tried it against the LARGE version.

[...]

    Just to add complications each release of the Tandem operating system
    produces a new version of the C compiler and the AXCEL program.  All
    of my objects are compiled under the "D30" operating system.

    Users running on different versions of NSK should (ideally) re-AXCEL the
    objects.  However, as is, the Zip/UnZip will probably run OK.

    The common versions of NSK that people will have are:

        C30     Zip/UnZip will NOT work, cannot even be compiled as they
                  use 'D' series calls
        D20     should work OK, re-AXCEL may help, re-compile should not
                  be necessary
        D30     will work 'as is'
        D40     should work OK, re-AXCEL may run faster, also re-compile
                  on D40 may run faster

    D40 is the current release of NSK, though most sites will be on D20/30.

Send problem reports on Zip and UnZip to:  http://www.info-zip.org/zip-bug.html
Send problem reports on gzip to:           support@gzip.org

Last updated:  12 January 2003