Bump version number prior to tagging 0.63.

[sgt/putty] / doc / pscp.but

\define{versionidpscp} \versionid $Id$

\#FIXME: Need examples

\C{pscp} Using \i{PSCP} to transfer files securely

\i{PSCP}, the PuTTY Secure Copy client, is a tool for \i{transferring files}
securely between computers using an SSH connection.

If you have an SSH-2 server, you might prefer PSFTP (see \k{psftp})
for interactive use. PSFTP does not in general work with SSH-1
servers, however.

\H{pscp-starting} Starting PSCP

PSCP is a command line application.  This means that you cannot just
double-click on its icon to run it and instead you have to bring up a
\i{console window}.  With Windows 95, 98, and ME, this is called an
\q{MS-DOS Prompt} and with Windows NT, 2000, and XP, it is called a
\q{Command Prompt}.  It should be available from the Programs section
of your \i{Start Menu}.

To start PSCP it will need either to be on your \i{\c{PATH}} or in your
current directory.  To add the directory containing PSCP to your
\c{PATH} environment variable, type into the console window:

\c set PATH=C:\path\to\putty\directory;%PATH%

This will only work for the lifetime of that particular console
window.  To set your \c{PATH} more permanently on Windows NT, 2000,
and XP, use the Environment tab of the System Control Panel.  On
Windows 95, 98, and ME, you will need to edit your \i\c{AUTOEXEC.BAT}
to include a \c{set} command like the one above.

\H{pscp-usage} PSCP Usage

Once you've got a console window to type into, you can just type
\c{pscp} on its own to bring up a usage message.  This tells you the
version of PSCP you're using, and gives you a brief summary of how to
use PSCP:

\c Z:\owendadmin>pscp
\c PuTTY Secure Copy client
\c Release 0.63
\c Usage: pscp [options] [user@]host:source target
\c        pscp [options] source [source...] [user@]host:target
\c        pscp [options] -ls [user@]host:filespec
\c Options:
\c   -V        print version information and exit
\c   -pgpfp    print PGP key fingerprints and exit
\c   -p        preserve file attributes
\c   -q        quiet, don't show statistics
\c   -r        copy directories recursively
\c   -v        show verbose messages
\c   -load sessname  Load settings from saved session
\c   -P port   connect to specified port
\c   -l user   connect with specified username
\c   -pw passw login with specified password
\c   -1 -2     force use of particular SSH protocol version
\c   -4 -6     force use of IPv4 or IPv6
\c   -C        enable compression
\c   -i key    private key file for authentication
\c   -noagent  disable use of Pageant
\c   -agent    enable use of Pageant
\c   -batch    disable all interactive prompts
\c   -unsafe   allow server-side wildcards (DANGEROUS)
\c   -sftp     force use of SFTP protocol
\c   -scp      force use of SCP protocol

(PSCP's interface is much like the Unix \c{scp} command, if you're
familiar with that.)

\S{pscp-usage-basics} The basics

To \I{receiving files}receive (a) file(s) from a remote server: 

\c pscp [options] [user@]host:source target

So to copy the file \c{/etc/hosts} from the server \c{example.com} as
user \c{fred} to the file \c{c:\\temp\\example-hosts.txt}, you would type:

\c pscp fred@example.com:/etc/hosts c:\temp\example-hosts.txt

To \I{sending files}send (a) file(s) to a remote server: 

\c pscp [options] source [source...] [user@]host:target

So to copy the local file \c{c:\\documents\\foo.txt} to the server
\c{example.com} as user \c{fred} to the file \c{/tmp/foo} you would
type:

\c pscp c:\documents\foo.txt fred@example.com:/tmp/foo

You can use \i{wildcards} to transfer multiple files in either
direction, like this:

\c pscp c:\documents\*.doc fred@example.com:docfiles
\c pscp fred@example.com:source/*.c c:\source

However, in the second case (using a wildcard for multiple remote
files) you may see a warning saying something like \q{warning:
remote host tried to write to a file called \cq{terminal.c} when we
requested a file called \cq{*.c}. If this is a wildcard, consider
upgrading to SSH-2 or using the \cq{-unsafe} option. Renaming of
this file has been disallowed}.

This is due to a \I{security risk}fundamental insecurity in the old-style
\i{SCP protocol}: the client sends the wildcard string (\c{*.c}) to the
server, and the server sends back a sequence of file names that
match the wildcard pattern. However, there is nothing to stop the
server sending back a \e{different} pattern and writing over one of
your other files: if you request \c{*.c}, the server might send back
the file name \c{AUTOEXEC.BAT} and install a virus for you. Since
the wildcard matching rules are decided by the server, the client
cannot reliably verify that the filenames sent back match the
pattern.

PSCP will attempt to use the newer \i{SFTP} protocol (part of SSH-2)
where possible, which does not suffer from this security flaw. If
you are talking to an SSH-2 server which supports SFTP, you will
never see this warning. (You can force use of the SFTP protocol,
if available, with \c{-sftp} - see \k{pscp-usage-options-backend}.)

If you really need to use a server-side wildcard with an SSH-1
server, you can use the \i\c{-unsafe} command line option with PSCP:

\c pscp -unsafe fred@example.com:source/*.c c:\source

This will suppress the warning message and the file transfer will
happen. However, you should be aware that by using this option you
are giving the server the ability to write to \e{any} file in the
target directory, so you should only use this option if you trust
the server administrator not to be malicious (and not to let the
server machine be cracked by malicious people). Alternatively, do
any such download in a newly created empty directory. (Even in
\q{unsafe} mode, PSCP will still protect you against the server
trying to get out of that directory using pathnames including
\cq{..}.)

\S2{pscp-usage-basics-user} \c{user}

The \i{login name} on the remote server. If this is omitted, and \c{host}
is a PuTTY saved session, PSCP will use any username specified by that 
saved session.  Otherwise, PSCP will attempt to use the local Windows
username.

\S2{pscp-usage-basics-host} \I{hostname}\c{host}

The name of the remote server, or the name of an existing PuTTY saved
session. In the latter case, the session's settings for hostname, port
number, cipher type and username will be used.

\S2{pscp-usage-basics-source} \c{source}

One or more source files. \ii{Wildcards} are allowed.  The syntax of
wildcards depends on the system to which they apply, so if you are
copying \e{from} a Windows system \e{to} a UNIX system, you should use 
Windows wildcard syntax (e.g. \c{*.*}), but if you are copying \e{from} 
a UNIX system \e{to} a Windows system, you would use the wildcard
syntax allowed by your UNIX shell (e.g. \c{*}).

If the source is a remote server and you do not specify a full
pathname (in UNIX, a pathname beginning with a \c{/} (slash)
character), what you specify as a source will be interpreted relative
to your \i{home directory} on the remote server.

\S2{pscp-usage-basics-target} \c{target}

The filename or directory to put the file(s).  When copying from a
remote server to a local host, you may wish simply to place the
file(s) in the current directory.  To do this, you should specify a
target of \c{.}.  For example:

\c pscp fred@example.com:/home/tom/.emacs .

...would copy \c{/home/tom/.emacs} on the remote server to the current 
directory.

As with the \c{source} parameter, if the target is on a remote server
and is not a full path name, it is interpreted relative to your home
directory on the remote server.

\S{pscp-usage-options} Options

PSCP accepts all the general command line options supported by the
PuTTY tools, except the ones which make no sense in a file transfer
utility. See \k{using-general-opts} for a description of these
options. (The ones not supported by PSCP are clearly marked.)

PSCP also supports some of its own options. The following sections
describe PSCP's specific command-line options.

\S2{pscp-usage-options-ls}\I{-ls-PSCP}\c{-ls} \I{listing files}list remote files

If the \c{-ls} option is given, no files are transferred; instead,
remote files are listed. Only a hostname specification and
optional remote file specification need be given. For example:

\c pscp -ls fred@example.com:dir1

The SCP protocol does not contain within itself a means of listing
files. If SCP is in use, this option therefore assumes that the
server responds appropriately to the command \c{ls\_-la};
this may not work with all servers.

If SFTP is in use, this option should work with all servers.

\S2{pscp-usage-options-p}\I{-p-PSCP}\c{-p} \i{preserve file attributes}

By default, files copied with PSCP are \i{timestamp}ed with the date and
time they were copied.  The \c{-p} option preserves the original
timestamp on copied files.

\S2{pscp-usage-options-q}\I{-q-PSCP}\c{-q} quiet, don't show \i{statistics}

By default, PSCP displays a meter displaying the progress of the
current transfer:

\c mibs.tar          |   168 kB |  84.0 kB/s | ETA: 00:00:13 |  13%

The fields in this display are (from left to right), filename, size
(in kilobytes) of file transferred so far, estimate of how fast the
file is being transferred (in kilobytes per second), estimated time
that the transfer will be complete, and percentage of the file so far
transferred.  The \c{-q} option to PSCP suppresses the printing of
these statistics.

\S2{pscp-usage-options-r}\I{-r-PSCP}\c{-r} copies directories \i{recursive}ly

By default, PSCP will only copy files.  Any directories you specify to
copy will be skipped, as will their contents.  The \c{-r} option tells
PSCP to descend into any directories you specify, and to copy them and 
their contents.  This allows you to use PSCP to transfer whole
directory structures between machines.

\S2{pscp-usage-options-batch}\I{-batch-PSCP}\c{-batch} avoid interactive prompts

If you use the \c{-batch} option, PSCP will never give an
interactive prompt while establishing the connection. If the
server's host key is invalid, for example (see \k{gs-hostkey}), then
the connection will simply be abandoned instead of asking you what
to do next.

This may help PSCP's behaviour when it is used in automated
scripts: using \c{-batch}, if something goes wrong at connection
time, the batch job will fail rather than hang.

\S2{pscp-usage-options-backend}\i\c{-sftp}, \i\c{-scp} force use of
particular protocol

As mentioned in \k{pscp-usage-basics}, there are two different file
transfer protocols in use with SSH. Despite its name, PSCP (like many
other ostensible \cw{scp} clients) can use either of these protocols.

The older \i{SCP protocol} does not have a written specification and
leaves a lot of detail to the server platform. \ii{Wildcards} are expanded
on the server. The simple design means that any wildcard specification
supported by the server platform (such as brace expansion) can be
used, but also leads to interoperability issues such as with filename
quoting (for instance, where filenames contain spaces), and also the
security issue described in \k{pscp-usage-basics}.

The newer \i{SFTP} protocol, which is usually associated with SSH-2
servers, is specified in a more platform independent way, and leaves
issues such as wildcard syntax up to the client. (PuTTY's SFTP
wildcard syntax is described in \k{psftp-wildcards}.) This makes it
more consistent across platforms, more suitable for scripting and
automation, and avoids security issues with wildcard matching.

Normally PSCP will attempt to use the SFTP protocol, and only fall
back to the SCP protocol if SFTP is not available on the server.

The \c{-scp} option forces PSCP to use the SCP protocol or quit.

The \c{-sftp} option forces PSCP to use the SFTP protocol or quit.
When this option is specified, PSCP looks harder for an SFTP server,
which may allow use of SFTP with SSH-1 depending on server setup.

\S{pscp-retval} \ii{Return value}

PSCP returns an \i\cw{ERRORLEVEL} of zero (success) only if the files
were correctly transferred. You can test for this in a \i{batch file},
using code such as this:

\c pscp file*.* user@hostname:
\c if errorlevel 1 echo There was an error

\S{pscp-pubkey} Using \i{public key authentication} with PSCP

Like PuTTY, PSCP can authenticate using a public key instead of a
password. There are three ways you can do this.

Firstly, PSCP can use PuTTY saved sessions in place of hostnames
(see \k{pscp-usage-basics-host}). So you would do this:

\b Run PuTTY, and create a PuTTY saved session (see
\k{config-saving}) which specifies your private key file (see
\k{config-ssh-privkey}). You will probably also want to specify a
username to log in as (see \k{config-username}).

\b In PSCP, you can now use the name of the session instead of a
hostname: type \c{pscp sessionname:file localfile}, where
\c{sessionname} is replaced by the name of your saved session.

Secondly, you can supply the name of a private key file on the command
line, with the \c{-i} option. See \k{using-cmdline-identity} for more
information.

Thirdly, PSCP will attempt to authenticate using Pageant if Pageant
is running (see \k{pageant}). So you would do this:

\b Ensure Pageant is running, and has your private key stored in it.

\b Specify a user and host name to PSCP as normal. PSCP will
automatically detect Pageant and try to use the keys within it.

For more general information on public-key authentication, see
\k{pubkey}.