Tips and techniques for exchanging files between Unix and DOS, MacOS, or VMS systems with FTP

By Becca Thomas

Long gone are the days when a computing shop used all the same systems. Now, sprinkled among the high-powered workstations and VMS-based minis on your network are various DOS and Windows PCs and Apple Macintoshes. The trick is to allow easy exchange of files between the divergent systems.

TCP/IP protocols not only allow you to copy files between different Unix system variants, but also between Unix and other operating systems, such as DOS, MacOS, and VMS, among others. This month we look at some tips and techniques for moving files between some of these operating-system environments with the file transfer protocol (FTP).

Dear Dr. Thomas:

One of the common problems in computing is transferring files between different platforms, say between a Unix-based workstation and either a PC running DOS or a VMS-based VAX minicomputer. Although FTP has made such transfers possible, you still need to know about the pitfalls associated with different file-naming schemes.

DOS <-> Unix

[Doctor's notes: DOS files are named with a disk- device designator, which is a single letter followed by a colon, then the file name with up to eight characters, and lastly an optional extension, consisting of a dot and up to three characters (also known as the ``8+3 rule''). Listing 1A shows this format. The backslash (\) character separates directories and file names in a DOS path name, as in C:\U\WERNER or its equivalent c:\u\werner. Letter case is not significant. Generally, DOS-based FTP servers translate case- sensitive Unix file names into case-insensitive DOS names during file transfers. These servers also have to truncate longer Unix file names to fit the 8+3 rule.]

Always use ``binary'' mode to transfer files containing control characters and eight-bit data such as compiled programs, compressed data, or other nontext data files. Use ASCII mode for text files, like Postscript programs. The DOS server should also translate carriage-return, line-feed terminators used in DOS text files to use the Unix newline terminator and vice versa.

Listing 1B shows that the FTP dir command yields a directory listing like that obtained with the DOS DIR command. Here, we log into the DOS server on system da2, run the FTP pwd command to display the path name of the current directory (the \U subdirectory on the ``C'' hard- drive volume), then list that directory.

To request files from another volume, use the FTP cd command to select that volume. For instance, enter cd a: to select diskette drive A. The DOS server will change the working directory to the diskette drive. Just be sure to have a diskette inserted in the drive before selection or the DOS server might crash. If it does, try to start up the server or reboot the system if it is hung.

MacOS <-> Unix

Unlike DOS, the Macintosh operating system doesn't separate the disk volume designator through a special syntax. A disk has a name like a directory or a file where the components are separated by colons (:) Listing 2A shows the general format. Here's a specific example:

Aktenschrank:The Beauty of Fractals Lab:Old Demos:Demo1

Unfortunately, when transferring MacOS binaries to some other architecture, the files will be corrupted even if you set the transfer mode to binary. You have to set ``MacBinary mode'' to transfer these files successfully. There should be a corresponding option on any file-transfer program running under MacOS, such as NCSA's (National Center of Supercomputing Applications) public-domain Telnet/FTP implementation.

MacOS file names can include punctuation characters and even spaces, which present a problem. Enclose whitespace characters in quotes and rename MacOS files--whose names include shell metacharacters--to legal Unix file names. Listing 2B shows a sample interactive file-transfer session.

VMS <-> Unix

[Doctor's notes: Listing 3A shows the general format of a VMS file name: the node name is a one- to six-character identifier of a computer system in a DECnet network; the four- to seven-character device name indicates the physical location where the file is stored, such as magnetic or optical disk; the directory name is a 1- to 39-character identifier that names the directory containing the file; the 1- to 39-character extension is generally used to indicate the type of file; and the version number is an integer in the range between 1 and 32,767.]

For file transfers you'll generally work with VMS files named with a disk designator, directory name, file name, optional file-extension name, and version number. For instance, DUA0:[ROHDATEN]ZSSN.RMB;1 denotes the first version of the file named ZSSN with extension RMB on the DUA0 disk in directory ROHDATEN.

Several standard extensions are used to denote the type of file, such as DIR for directory files, TXT for human-readable text, EXE for executables, and COM for command files. Although VMS maintains previous versions of a file, hence the version number, you can ignore it and VMS will hand you the latest version or you can specify an earlier version by number, but be sure to rename the file if it's destined for a Unix system so the semicolon doesn't become part of the file name. Also, you should remove or translate other shell metacharacters, such as the square brackets.

Part B of Listing 3 shows a sample directory listing. The numbers separated by a slash (/) denote the file size: the first is the number of 512-byte blocks used out of the total allocated, which is specified by the second number. All the files shown have used 17 out of 18 blocks. Next the creation date and time is displayed. Note that the details of your display may differ, such as the order of the fields. Or perhaps it has additional information, such as the file-access modes.

Use the FTP cd command to change to another directory. VMS directories look like any other file except for the .DIR file-name extension. However, you'll omit that extension when changing directories. Unfortunately, there are two distinct ways to represent multiple directory levels. One uses the Unix ``slash'' notation, as in a/b/c, but the other uses dots as in [a.b.c]. Many servers don't recognize dot-dot (..) as the parent directory, so you'll need to use cd [.-], cd [-], or the cdup command to move up one directory level.

If you're using the dot-style syntax, be sure to include the dot for a subdirectory of your current directory, as in cd [.filetrans], or the server assumes filetrans is in the top-level (root) directory for the current disk. Letter case isn't significant, so you could enter cd [.FILETRANS] or even cd [.FiLeTrAnS] to move down one level to the FILETRANS.DIR subdirectory. Listing 3C shows some examples.

FTP servers provide the ntrans command that lets you specify translation of characters in file names when copying files. Listing 3D shows an example. Here we translate the first character in the input-character string (;[]:) to a dot, but remove the others by translating them to nothing because there's no corresponding output character specified. The subsequent file-retrieval request shows that DUA0:[ROHDATEN]ZSSN.RMB;1 is translated to DUA0ROHDATENZSSN.RMB.1.

It's crucial to transfer ASCII files in mode ASCII because VMS files have more attributes than Unix files, so you might end up with your file being truncated, missing the line-terminator characters, or having some other corruption.

Proxy Transfers

If you want to transfer files between two remote hosts, you can use the proxy command, which comes in handy when your local network is behind a firewall machine. You could log onto your account on the firewall and then use the FTP proxy command to copy files from the Internet onto your machine behind the firewall. When you ``open'' a connection, it will be the primary connection to your commands' destination. If you then open a proxy connection with proxy open other.host, you have a virtual connection between your primary host and other.host. If you tell the proxy host to get a file from the primary host, you have to type in proxy get filename as shown [in Listing 4]. You can use any FTP command the proxy host understands.

Tester's Comments

• NT uses drive letters similar to DOS, but the NT file system itself looks more like Unix. If you use NT with the DOS file system (the Microsoft code name is FAT file system), the server uses case-insensitive 8+3 names.
  
  Use the quote site dirstyle command to switch the directory listing (dir command) output format between Unix or DOS styles. The default format is configuration dependent. Note that the ftp.microsoft.com host (also known as gowinnt.microsoft.com) uses the Unix ls -l listing style when you log in, whereas DOS style is used for an out-of-the-box system. The quote site ckm command toggles the cd command's verbose mode.
  
  Novell Netware uses volumes instead of drive letters. Normally, all Netware servers have a SYS volume, but that may be unreadable. Additional volumes are named VOLN by default, but may be renamed by the administrator. Case-insensitive 8+3 DOS file names are used by default, but NFS or MacOS name spaces are available.
  
  When the local system is DOS, Windows NT with DOS file system, or System V before Release 4, don't forget to rename long names when retrieving files from remote BSD, SVR4, Windows NT with NT file system, or MacOS file systems. Otherwise the file name will be truncated without notice, which may cause overwriting of files using multiple ``get'' (mget) commands.
  
  Novell sells FTP servers in two packages: Netware NFS and Netware FLEXIP. I began setup of the FTP server for the NFS product. The FTP server (FTPSERV.NLM) is inetd-daemon based (INETD.NLM).
  
  Although Netware normally uses DOS-style file names, it can support Unix-like (NFS.NAM), MAC-like (MAC.NAM), and OS2-like (OS2.NAM) file names. We don't use the Unix-compatible names at our facility because it slows down the mounting of the Netware volumes significantly.--Endre Bálint Nagy, Walton Networking Ltd., Budapest, Hungary
  
  
• I browsed through the header files of a few Unix utilities used to unpack MacOS archives, and there doesn't seem to be file-name-length consensus: I've seen 32-bytes, 64-bytes, and 255-bytes defines.
  
  You can change Netware volumes with the change directory command, as in:

  cd volumename:/dirname
  

  much like it works with DOS, except you have to type a name instead of one letter. Also, there's no way to find out the volume names if you don't know them. My guess is that it holds true for VMS devices and MacOS disk names also because there is no single ``root'' directory at the top of the hierarchy. Unlike Unix, the Netware FTP daemon displays permissions you have.
  
  Some tests I ran included logging in from an SCO Unix ODT 3.0 system to a PC running Desqview/X on top of DOS 6.2, using Desqview's FTP daemon. All went well, even connecting back to the ODT box using proxy. Then, I logged in from the same Unix box to a Novell Netware 3.11 server, using Murkworks Inc.'s FTP server (NLM-based). The file names look similar to DOS file names but have a volume name prefixed instead of a one-letter device name. Murkworks' FTPD.NLM also supports proxy connections.--Kees Hendrikse, Echelon Consultancy, Enschede, The Netherlands
  
  
• Most of our campus users employ DOS/Windows FTP clients that automatically deal with Unix file names (truncating them as a rule). Few people allow FTP into their PC because they're shut off at night. More importantly, some packages don't require a password to access the FTP server, although that can be configured.--Mike Peterson, System Administrator, Department of Chemistry, University of Toronto, Canada.