Linux 版 (精华区)
发信人: netiscpu (说不如做), 信区: Linux
标 题: [B] Red Hat Linux Unleashed (43)
发信站: 紫 丁 香 (Sat Jul 25 05:02:13 1998), 转信
UUCP
_________________________________________________________________
o UUCP Configuration
# Taylor UUCP Configuration
# HDB UUCP Configuration
o A UUCP Connection
o Direct Connections
o Login Scripts
o Changing Access Times
o UUCP Security
o Using UUCP
# Sending E-Mail with UUCP
# Transferring Files with UUCP
# Checking on Transfers
o Summary
_________________________________________________________________
43
UUCP
UUCP (UNIX to UNIX CoPy) was developed to provide a simple dial-up
networking protocol for UNIX systems. It is most often used today as
an e-mail transfer system, allowing non-networked machines to transfer
e-mail easily over a modem connection. It can also be used for Usenet
news and access to similar services that do not require a dedicated
connection. UUCP is a two-machine connection, between your Linux
machine and another machine running UUCP. UUCP cannot be used as a
remote system access system (like FTP or Telnet), nor can it be used
as a standard login because the protocols do not support this type of
interactive behavior. UUCP does have security features that are
adequate for most purposes, but UUCP is the primary source of
break-ins for many systems because most system administrators don't
bother to set the security system up properly.
Linux can run any of several different versions of UUCP, most of which
are compatible with each other to a reasonable extent, except when it
comes to configuration and installation procedures. Many Linux
versions offer you a choice between the Taylor UUCP version (included
on the Red Hat distribution accompanying this book) and the HDB
(HoneyDanBer) UUCP. You can use whichever single version came with
your Linux software, or if you have both, you can choose between the
two (or use both versions as the mood strikes you). Many Linux users
prefer the Taylor UUCP implementation, while users who have worked on
other UNIX systems prefer HDB because it is more recent. We'll look at
both versions in this chapter. (There are even more UUCP versions, but
we will ignore them because they are seldom used under Linux.) The
first part of the chapter deals with configuring UUCP, while the rest
covers using it.
UUCP Configuration
Most of the configuration required for UUCP takes place under the
/usr/lib/uucp directory. There are several files used by UUCP, most of
which need direct administrator modification to set up properly. While
the configuration process can seem awfully complex to someone who has
never done it before, there are really only a few files that need
changing, and only one or two entries in each file.
The configuration process for Taylor UUCP and HDB UUCP are completely
different, so we will look at them separately. However, you don't have
to worry which version of UUCP is being run at the remote end of the
connection because both can talk to each other (at least that's
usually the case), as long as the configuration files are set up
properly.
Some versions of Linux have semiautomated UUCP configuration scripts.
These are more common with HDB UUCP than Taylor UUCP, but a few
helpful scripts are also available for the latter. If you have one of
these scripts, by all means use it, but do check the files manually
afterwards.
For the configuration processes discussed in the following section, we
will assume that our host machine's name is merlin, and we want to
connect via UUCP to another Linux system called arthur. As you go
through the process, take care to enter the information in the same
format as the examples, but don't mix Taylor and HDB UUCP information.
Taylor UUCP Configuration
We can begin with a quick look at the configuration files involved in
the Taylor UUCP system. These are the filenames and their primary
purposes:
* /usr/lib/uucp/config—Defines the local machine name.
* /usr/lib/uucp/sys—Defines the remote systems and how to call
them.
* /usr/lib/uucp/port—Describes each port for calling out and
its parameters.
* /usr/lib/uucp/dial—Describes the dialers for calling out.
* /usr/lib/uucp/dialcode—Used to contain expansions for
symbolic dialcodes, but is rarely used when a straight-out
telephone connection exists.
* /usr/lib/uucp/call—Can contain the login name and password
for remote systems, but it is rarely used now.
* /usr/lib/uucp/passwd—Contains the login names and passwords
used when remote systems connect to your local machine. Used only
when uucico is password checking instead of using the login
process.
To make the configuration process easier, we will ignore all the
theory and background information and proceed with a sample
configuration. You need only modify the entries to suit your own
names, telephone numbers, device files, and so on, and the
configuration process will be the same. It can then be repeated for as
many systems as you want to connect to.
The first file to modify holds your system name and other general
parameters. The file /usr/lib/uucp/config needs a single line entry
for your system name such as this one:
nodename merlin
The keyword nodename must be first on the line followed by whitespace
(spaces or tabs), then your machine name. The information in this file
may have been completed when you installed Linux, but you should
manually check the contents to make sure. If your system's name isn't
set correctly, the connection to the remote system won't work
properly.
______________________________________________________________
NOTE: To use UUCP, you must have a system name. For compatibility
with most versions of UUCP, keep the name to seven characters or
less. Ideally, the UUCP name is the same name you assigned to your
host during configuration. The name doesn't have to follow a
convention (such as the system name used by TCP/IP for Internet
access), but if you use other network protocols, keep a consistent
name. If you have a domain name (for TCP/IP access) use the first
component of the machine's full TCP/IP name as the UUCP name. For
example, if your full domain name is merlin.wizards.com, use the
UUCP name merlin.
______________________________________________________________
You also need information about the remote system you want to connect
to. The /usr/lib/uucp/sys file holds all the information about remote
systems. This file usually has a few sample entries in it, which you
can copy or modify. Make sure you don't leave comment marks (pound or
hash marks) in the first column or the entries will be ignored. A
/usr/lib/uucp/sys entry for the remote machine merlin looks like this:
# system: arthur (Bill Smallwood's Linux system)
system arthur
time Any
phone 555-1212
port com1
speed 9600
chat login: merlin password: secret1
The first line in the preceding extract is a comment line. Most system
administrators like to put a comment line in to identify each system.
The next lines identify the different aspects of the remote system,
including its name (arthur), times at which it can be called (Any in
this case, meaning no restrictions), the telephone number (including
any area code or special digits that have to be dialed), the serial
port to be used for the connection (in this case, com1), the speed at
which to connect (9600 baud), and the chat script or login process. In
this case, the chat script tells UUCP to wait until it sees the string
login: then to send merlin, then wait for the prompt password: and
then to send secret1.
Most login scripts will require a login and password, and they must be
placed in the configuration file because UUCP doesn't allow
interactive sessions. This can be a bit of a problem because it allows
other users on your system to see the login password for the remote
machine, but since it can only be used by UUCP, this is not a major
concern. Also, the file permissions on the UUCP configuration files
can be set to prevent any system users (other than root) from looking
into the file.
______________________________________________________________
NOTE: Not all remote sites need a password for entry through UUCP.
For example, some public archives let you log in and retrieve files
using the uucp login with no password. Some sites use readily
available passwords, such as uucp.
______________________________________________________________
The port name used in the /usr/lib/uucp/sys entry does not have to
match a device name on the Linux system, because another file is used
to match the entry to a physical device. This file is
/usr/lib/uucp/port, and it requires an entry similar to this for a
9600 baud modem:
# com1 device port
port com1
type modem
device /dev/cua0
speed 9600
dialer Hayes
In the /usr/lib/uucp/port file, the name of the port used in the
/usr/lib/uucp/sys file is identified on the first line. The type of
connection to be used (usually modem) is on the next. The actual Linux
device that corresponds to the port name is specified as a device
driver (for many Linux systems this can be /dev/modem, which is linked
to the serial port device driver).
The modem connection speed comes next, and shows the maximum speed the
modem can be used at. Finally, the name of a dialer is entered. This
is a throwback to the days when modems couldn't dial themselves, but
used another device (called a dialer) to make the connection.
The dialer entry in the /usr/lib/uucp/port file is then matched to an
entry in the file /usr/lib/uucp/dial, which tells the modem how to
dial the phone. Here's a simple entry:
# Hayes modem
dialer Hayes
chat "" ATZ OK ATDT\T CONNECT
This shows the script that the system uses to communicate to the Hayes
modem. In this case, the \T in the command line is replaced with the
telephone number to be called. Some Linux systems simplify the use of
the /usr/lib/uucp/port and /usr/lib/uucp/dial files into one single
entry in the /usr/lib/uucp/sys file, which names the modem file
directly.
The remote end of the connection (in this case, the system arthur)
must have corresponding entries for merlin. The files will be similar
with only name, telephone number, and (possibly) device name and chat
script changes. Until both ends are configured properly, you can't get
a connection between the two machines.
Some Linux systems with Taylor UUCP have a utility called uuchck that
verifies the syntax in the UUCP configuration files and prints out
summary information. If you don't have the uuchck utility, it can be
downloaded from many FTP and BBS sites. Check Appendix A for some of
the possible sites.
By default, Taylor UUCP allows a remote system to execute only a
limited number of commands when they log into your system. Typically,
the remote is only allowed to execute rmail and rnews, to transfer
mail and news, respectively. If you want to allow extra programs to be
executed, add a line to the /usr/lib/uucp/sys file that includes all
the commands the remote system can execute. For example, the entry:
system chatton
....
commands rmail rnews rsmtp rdataupdate
specifies that the system chatton can execute any of the four commands
given after the commands keyword. Note that all four commands must be
in the usual search path used by the UUCP utilities (actually by
uuxqt).
If you intend to transfer files between two machines, you must also
modify the configuration files. When a remote system sends a file to
your machine, the files usually should be stored in the directory
/usr/spool/uucppublic (some systems use /var/spool/uucppublic) as a
safety precaution. You don't want to allow a remote system to write
files anywhere on your file system, or they could overwrite critical
system files. The convention for most UUCP systems is to use either
/usr/spool/uucppublic or /usr/spool/uucp/system (where system is the
remote system's name) as the transfer directories.
You can specify transfer and receive directories in the
/usr/lib/uucp/sys file. For example, the following entry for the
remote system chatton has been modified to include specific
directories for file transfers:
system chatton
...
local-send ~/send
local-receive ~/receive
In this configuration, the users on your local machine can send any
file that is in the send directory under the uucp directory (~/send,
which means that any file to be sent to a remote system must be
transferred there first), and any file incoming from a remote system
is stored in the receive directory under the uucp directory. If you
want to allow transfers from a user's home directory, you can specify
the /usr directory as a starting point. Multiple entries are separated
by spaces, so the entry
local-send ~/send /usr
allows transfers from the send directory under the uucp directory, or
from any directory under /usr.
The preceding two lines deal only with file transfers requested or
sent from your machine. If you want to enable requests for transfers
from the remote machine, you need to add two more lines:
remote-send /usr/lib/uucppublic
remote-request /usr/lib/uucppublic
This forces the remote machine to request files and send them only to
the /usr/lib/uucppublic directory. Again, you can offer several
choices if you want, as long as they are separated by spaces.
Finally, UUCP allows machines to forward data through other machines,
a process called hopping. In other words, if you want to send mail to
the system warlock but can only get there through the system wizard,
you have to instruct UUCP that your local system can get to warlock
through wizard. You add a forward command to the /usr/lib/uucp/sys
file:
system wizard
...
forward warlock
You should then add an entry for the warlock system that tells UUCP
that any mail for you will be coming back through wizard. The matching
entry would be:
system warlock
...
forward-to merlin
The forward-to entry is necessary so that any files returned by
warlock are passed to merlin, the local host machine. Otherwise, they
would be discarded by UUCP as not being routable.
By default, Taylor UUCP does not allow forwarding, and most system
administrators should think carefully about allowing it because the
potential for abuse is high.
HDB UUCP Configuration
HDB UUCP is a more recent version of UUCP and its configuration files
are different. In many ways, the HDB configuration is easier than
Taylor UUCP's, although neither is difficult once you know the basic
process.
The name of the local system is not set in the UUCP configuration
files but by the Linux hostname itself. To set the system name use the
hostname command.
The names of the remote systems are stored in the file
/usr/lib/uucp/Systems (some older versions used the name
/usr/lib/uucp/L.sys). There is a single line for each remote system
that you will be connected to. The format of each line is as follows:
sitename schedule device_type speed phone login_script
where sitename is the name of the remote machine, schedule is when it
can be connected to, device_type is the type of device to use to call
the remote system, speed is the speed (or range of speeds) that can be
used to connect, phone is the telephone number, and login_script is
the script used when a connection is made (such as the chat script in
Taylor UUCP). For example, to call the remote system arthur, the
/usr/lib/uucp/Systems file would have a line like this:
arthur Any ACU 9600 555-1212 login: uucp password: secret1
The Any entry in the schedule field tells UUCP that it can call at any
time. The ACU entry in the device field tells UUCP to use the ACU
(automatic calling unit) defined in the /usr/lib/uucp/Devices file.
The /usr/lib/uucp/Devices file (or /usr/lib/uucp/L-devices file in
some older versions) contains information about the devices that can
be used to call the remote systems. The Devices file follows this
syntax:
devicetype ttyline dialerline speed dialer [token Dialer ...]
where devicetype is the name of the device (which should match the
device name in the /usr/lib/uucp/Systems file), ttyline is the device
driver to be used for the connecting port (usually a serial line, such
as /dev/tty2a or /dev/modem), dialerline is an obsolete field left as
a hyphen, speed is the speed range of the device, and dialer is the
name of the file that tells UUCP how to use the device. A sample line
for a Hayes 9600 baud modem used to connect on the second serial port
of the system might have an entry in the /usr/lib/uucp/Devices file
like this:
ACU tty2A - 9600 dialHA96
This identifies the ACU entry as a 9600 baud connection through
/dev/tty2A (the /dev portion of the name is not needed with HDB UUCP),
and it uses a program called dialHA96 to handle the setup and dialing
of the modem. There are usually programs available for most popular
modems that set the modem configuration parameters automatically,
leaving Linux out of that process.
If a modem program is not available to handle the modem, an entry in
the file /usr/lib/uucp/Dialers can be used. The format of the Dialers
entries is:
dialer translation expect send ...
where dialer is the name of the dialer (matching the Devices file),
translation is the translation table to use for the phone number
(converting characters where needed to pauses, beeps, and so on), and
the expect/send entries are the chat script to set up the modem. A
sample line in the Dialers file looks like this:
hayes1200 =,-, "" AT\r\c OK\r \EATDT\T\r\c CONNECT
This is the entry for a Hayes 1200 Smartmodem, identified by the name
hayes1200, with translations for the = and - characters, followed by
the AT commands used to set up the modem. Since these entries are
usually supplied in the Dialers file for most popular modems, we won't
bother going into detail about them.
Permissions for file transfers are a little more convoluted with HDB
UUCP than Taylor UUCP, because HDB UUCP adds many features for special
handling. Instead of looking at all the file transfer permissions, you
are better advised to consult a specialty book on UUCP, because the
subject can easily consume 100 pages by itself! For this reason, we'll
look at the fundamentals—just enough to get you set up properly.
Permissions for remote system access and file transfers are handled by
the file /usr/lib/uucp/Permissions. The general format of the entries
in this file is:
MACHINE=remotename LOGNAME=uucp \
COMMANDS=rmail:rnews:uucp \
READ=/usr/spool/uucppublic:/usr/tmp \
WRITE=/usr/spool/uucppublic:/usr/tmp \
SENDFILES=yes REQUEST=no
where MACHINE identifies the remote machine's name, LOGNAME is the
name they use to log in (or you use to log in to their system),
COMMANDS are the commands they can execute on your local system, READ
is the list of directories they can read files from, WRITE is the list
of directories where they can write files, SENDFILES means they can
send files (yes or no), and REQUEST means they can request files from
your system (yes or no). Notice the use of slashes at the end of the
first four lines, to indicate this is really a single long line broken
up for readability. This is a typical UNIX convention.
A complete entry for the remote system wizard shows that it is allowed
to both send and receive files, but only from the
/usr/spool/uucppublic directory, and it can only execute mail and uucp
commands (the later transfers files):
MACHINE=wizard LOGNAME=uucp1 \
COMMANDS=rmail: uucp \
READ=/usr/spool/uucppublic: \
WRITE=/usr/spool/uucppublic: \
SENDFILES=yes REQUEST=yes
To prevent the remote system from sending files, change SENDFILES to
no. To prevent the remote system from requesting files, change REQUEST
to no.
A UUCP Connection
When UUCP connects to a remote machine, it follows a particular series
of steps. You can better understand the configuration files used by
UUCP, and the processes that are involved, by following through a
typical session. UUCP uses a process called uucico (UUCP Call In/Call
Out) to handle the process of connecting and sending information. A
UUCP connection can be started with the uucico command followed by the
remote system name, such as:
uucico -s arthur
When uucico starts, it examines the /usr/lib/uucp/sys file (Taylor
UUCP) or /usr/lib/uucp/Systems (HDB UUCP) to see if the remote system
name exists there. When it finds the proper remote system name, uucico
reads the rest of the entries for that system, including the port to
be used. From there, uucico uses /usr/lib/uucp/port and
/usr/lib/uucp/dial (Taylor UUCP) or /usr/lib/uucp/Devices and
/usr/lib/uucp/Dialers (HDB UUCP) to start the modem connection
(assuming it is a modem used to establish the session, of course).
When the modem is in use, uucico creates a lock on the device so no
other application can use it (the lock is a file starting with LCK..
and followed by the device name, such as LCK..cua0).
Once the chat scripts for setting up and dialing the modem have been
executed and the remote system is connected, uucico uses the chat
script in the /usr/lib/uucp/sys or /usr/lib/uucp/Systems file to log
into the remote. Once logged in, the remote machine starts up its copy
of uucico, and the two uucico processes establish handshaking between
themselves. Finally, after the handshaking has been established,
uucico goes ahead and handles any transfers that are queued.
When completed with the session, the local machine checks to make sure
that the remote has nothing further to send and then breaks the
connection. Finally, uucico terminates.
Direct Connections
If your two machines are directly connected through a serial port, for
example (no modems involved in the connection), you can use UUCP as a
simple network protocol for file transfer. The only changes to the
configuration files mentioned earlier are in the port specification.
Instead of using a modem device, you specify a direct connection. For
example, in the /usr/lib/uucp/sys file (Taylor UUCP), you would have
an entry like this:
port direct1
with a matching entry in the /usr/lib/uucp/port file that looks like
this:
port direct1
type direct
speed 38400
device /dev/cua1
with the speed of the direct connection and the port which uses it
specified. The entries in the HDB UUCP version are similar, using the
/usr/lib/uucp/Systems and /usr/lib/uucp/Devices files.
Login Scripts
The login scripts that form part of the /usr/lib/uucp/sys or
/usr/lib/uucp/Systems file can be the most difficult part of a UUCP
connection to get correct. If the machine you are logging into is a
typical UNIX system, there should be only the usual login and password
prompts to worry about. Other systems may require some special
handling to gain access. For this reason, the login script is worth a
quick look.
Generally, the layout of the login script is in a pattern-action pair,
with the pattern coming from the remote machine and the action from
the local. A simple login is:
login: merlin password: secret1
In this case, the local system waits until it sees the string login:
coming from the remote, sends merlin, waits for password:, then sends
secret1. You can simplify the script a little by cutting out extra
letters from the remote, because all you really need are the last
couple of characters and the colon. The script could have been written
like this:
gin: merlin word: secret1
This type of script has a good use. Suppose the remote system sends
Login: instead of login:; then the shortened form will work and the
longer match won't.
One useful feature of the uucico login script is the ability to wait
for the remote machine to reset itself (or start a getty process, more
likely). This is implemented by using a hyphen and the word BREAK in
the script, which tells uucico to send a break sequence if the remote
site hasn't responded in a timely manner. For example, the script
would be similar to this:
ogin:-BREAK-ogin: merlin sword: secret1
In this case, if the remote machine doesn't respond with an ogin:
prompt after a short period of time, the local machine sends a break
sequence and waits for the prompt again.
A few special characters can be used in the login script. The most
important ones for most UUCP purposes are the following:
\c Suppresses sending carriage return (send only)
\d Delays one second (send only)
\p Pauses for a fraction of a second (send only)
\t Sends a tab (send and receive)
\r Sends a carriage return (send and receive)
\s Sends a space (send and receive)
\n Sends a newline (send and receive)
\\ Sends a backslash (send and receive)
Sometimes you need to use one or more of the characters to get the
remote machine to respond to a modem login. For example, the following
script
\n\r\p ogin: merlin word: secret1
sends a carriage return-line feed pair before starting to match
characters. This is usually enough to get the remote machine to start
a getty on the port.
Changing Access Times
Both Taylor and HDB UUCP versions let you specify a time to call the
remote systems. While the examples so far show Any (meaning the system
can be called at any time, day or night), you may want to restrict
calls to local-cost times, or on certain days of the week. The reason
for limiting calls may be at your end (costs, for example), or at the
remote (limited access times during the day, for example).
To specify particular days of the week to allow calls, use a two-digit
abbreviation of the day (Mo, Tu, We, Th, Fr, Sa, Su), or Wk for
weekdays (Monday through Friday), Any (for any time), or Never (for
not allowed to connect). Any combination of the days may be used, as
you will see in a moment. The times for connecting are specified as a
range, in 24-hour format, when a time span is required. If no time is
given, it is assumed that any time during the day is allowed.
Dates and times are run together without spaces, while subsequent
entries are separated by commas. Examples of restricted access times
are as follows:
Wk1800-0730
MoWeFi
Wk2300-2400, SaSu
The first example allows connection only on weekdays between 6 p.m.
and 7:30 a.m. The second allows connection any time on Monday,
Wednesday, or Friday. The last example allows connections only between
11 p.m. and midnight on weekdays, and any time on weekends. You can
build up any time and date specifications you want. These apply to
both Taylor and HDB UUCP versions.
UUCP Security
The permissions of the UUCP configuration files must be carefully set
to allow UUCP to function properly, as well as to allow better
security for the system. Simply stated, the files should all be owned
by uucp, and the group should be uucp on most systems that have that
group in the /etc/group file. The ownerships can be set either by
making all the file changes while logged in as uucp, or by setting the
changes as root and then issuing the following commands when you are
in the /usr/lib/uucp directory:
chown uucp *
chgrp uucp *
As a security precaution, you should set a strong password for the
uucp login if there is one on your system. Some versions of Linux do
not supply a password by default, leaving the system wide open for
anyone who can type uucp at the login prompt!
The file permissions should be set very tightly, preferably to
read-write (and execute for directories) only for the owner (uucp).
The group and other permissions should be blanked because a read
access can give valuable login information, as well as passwords to
someone.
When UUCP logs into a remote system it requires a password and login.
This information is contained in the /usr/lib/uucp/sys or
/usr/lib/uucp/Systems files, and should be protected to prevent
unauthorized snooping by setting file ownerships and permissions as
mentioned.
If you have several systems connecting into yours, they can all use
the same uucp login and password, or you can assign new logins and
passwords as you need them. All you need to do is create a new
/etc/passwd entry for each login (with a different login name from
uucp, such as uucp1, uucp_arthur, and so forth) and a unique passwd.
The remote system can then use that login to access your system. When
you create new UUCP users in the /etc/passwd, force them to use uucico
only to prevent access to other areas of your system. For example, the
login uucp1, shown here, forces uucico as the startup command:
uucp1::123:52:UUCP Login for Arthur:/usr/spool/uucppublic:/usr/lib/uucp/uucico
The home directory is set to the uucppublic directory, and uucico is
the only startup program that can be run. Using different logins for
remote machines also allows you to grant different access permissions
for each system, preventing unwanted access.
You should also carefully control the commands that remote systems can
execute on your local machine. This is done through the permissions
fields of the local access file and should be monitored carefully to
prevent abuse and unauthorized access. In a similar manner, if you are
allowing forwarding of files through your system, you should control
who is allowed to forward, and where they are forwarded to.
Most important of all is to ensure that whoever accesses your system
on a regular basis is someone you want access to be granted to. Don't
leave your system wide open for anyone to enter because you are
guaranteeing yourself disaster. Carefully watch logins and make sure
file permissions and ownerships are properly set at all times.
Using UUCP
Once you have configured UUCP, you can use it to transfer files and
e-mail. In order to use UUCP, you have to know the addressing syntax,
which is different than what you may know from the Internet. The UUCP
address syntax is:
machine!target
where machine is the remote machine name and target is the name of the
user or file that you are trying to get to. For example, to send mail
to the user yvonne on machine arthur, you would use the mail command
with the destination user name:
mail arthur!yvonne
UUCP lets you move through several machines to get to a target. This
can help save on telephone bills, or make a much wider network
available to you from a small number of connections. Suppose you want
to send mail to a user called bill on a system called warlock, which
isn't in your configuration files but can be connected to through
arthur. If you have permission to send mail through the system arthur
(called a "hop"), you could send the mail with this command:
mail arthur!warlock!bill
When UUCP decodes this address, it reads the first system name
(arthur) and sends it to that system. The UUCP processes on arthur
then examines the rest of the address and realizes the mail is to be
sent on to warlock. If you have permission to forward through arthur,
UUCP on arthur sends the mail through to warlock for you. You can have
many hops in an address, as long as each system you are connecting to
allows the pass-through and can connect to the next machine on the
list. For example, the address:
arthur!warlock!chatton!vader!alex
would send data through arthur, warlock, chatton, and vader in order,
and then to the user alex. The addresses must be specified in the
proper hop order or the address will fail. This multihop addressing
can be very useful if a number of friends have local connections to
other machines, allowing you to easily set up a complex network. The
hard part is usually tracking the names of the systems involved. (The
exclamation mark in the address is called a bang, so the address above
is spoken as "arthur-bang-warlock-bang-chatton-bang-vader-bang-alex.")
______________________________________________________________
NOTE: Some shells don't like the bang character because they are
interpreted as a special shell command. Shells, such as the C
Shell, use the exclamation mark to recall previous commands, so you
must escape the bang with a slash to prevent the shell's
interpretation. Addresses then become arthur\!chatton\!yvonne. This
looks funny, but you get used to it.
______________________________________________________________
Depending on how you have your UUCP system set, it may call out to the
other systems in an address whenever something is submitted to it, or
if callout times are limited, the data may be spooled until a call is
allowed. You have already seen how to set callout times in the
/usr/lib/uucp/sys and /usr/lib/uucp/Systems files.
A quick caution about relying on UUCP for delivery of information: If
the systems that are being used are not set to call immediately when
there is something queued, your data can take a long time to get to
its destination. For example, if one of the hops in your address calls
the next machine in the address only once a day, you may have a
24-hour delay in delivery. This can be exacerbated by each machine in
the network.
______________________________________________________________
NOTE: Don't rely on the contents of your data sent through UUCP to
be kept confidential. Once on a remote system, any user with access
privileges to the queue could snoop into your data. Ideally, the
file permission prevents anyone but the superuser accessing the
data, but not all systems keep tight security. If you must send
sensitive data, encrypt it and let the recipient know the
decryption key through another format (not in a mail message!).
______________________________________________________________
UUCP deals with all transfers as jobs, a term you'll encounter often
when working with UUCP and its documentation. A job is a command that
is to be executed on the remote system, a file that is to be
transferred to or from the remote, or any other task that you want
performed between the two systems.
Sending E-Mail with UUCP
Most utilities like mail packages understand the UUCP addresses, so
that you don't have to worry about e-mail not reaching the proper
destination. You don't usually have to make any changes at all to
applications running under Linux to get them to understand this UUCP
address format. In the last section you saw how the mail package can
be used with UUCP addresses.
You can use any of the usual mail command options to modify the
behavior of the package. For example, to send the contents of the file
data_1 to yvonne on system chatton through the system arthur, and tag
the mail with a subject heading, you would issue the command:
mail -s "Data file" arthur\!chatton\!yvonne < data_1
Most mail packages available for Linux, including X-based mailers,
work perfectly well with UUCP addresses, as well as the more common
Internet addresses, but you may want to check before adopting a new
mail package.
Transferring Files with UUCP
UUCP's most common use is to transfer files from one machine to
another. To transfer files using UUCP, you use the uucp command. The
syntax of the uucp command is:
uucp [options] source destination
The options supported by UUCP vary a little depending on the version
and type of UUCP implementation, but most versions support the
following useful options:
-c Do not copy the file to a spool directory before sending. The
default action is to copy to a spool directory, and using the -c
option explicitly specifies it.
-f Don't create directories on the remote system, if needed. The
default action is to create directories as needed and can be
explicitly specified with the -d option.
-m Sends mail to the person who issued the uucp command when the copy
is complete.
-n user Sends mail to the user on the remote system when the copy is
complete.
The default behaviors are usually sufficient for most users, although
you may want the mail options when you need confirmation of an action.
Both source and destination are the names of files or directories as
appropriate, much like the cp command. However, when you are dealing
with a remote system for the source or destination, you need to format
the file or directory in valid UUCP address formats. For example, to
send the file data_1 from your local machine's current directory to
the directory /usr/spool/uucppublic on the machine arthur, you would
use the command:
uucp data_1 arthur\!/usr/spool/uucppublic
Notice that the remote machine name was prepended to the full target
directory name. In most cases, when transferring files to remote
systems you will want to use the uucppublic directories as you likely
will not have permission to transfer files anywhere else in the file
system. Once on the remote system in the /usr/spool/uucppublic
directory, it is up to the remote system's users to find the file and
copy it to its intended destination directory.
If you want to send the same file to the user bill on the remote
machine, and store it in a subdirectory called
/usr/spool/uucppublic/bill and send mail to both yourself and bill
when the copy is completed, you would issue the command:
uucp -m -nbill data_1 arthur\!/usr/spool/uucppublic/bill/
To copy a file from a remote machine to yours, you need to specify the
location of the remote. Remember, you must have access to the
directory that the files reside in (as well as read permission on the
file), or have the sender copy them to uucppublic. The command
uucp chatton\!/usr/tmp/bigfile /usr/tparker/
would transfer the bigfile file from the directory /usr/tmp on the
machine chatton to your /usr/tparker directory.
UUCP allows wildcards to be used, although you must escape them in
quotation marks to prevent the shell from interpreting them. For
example, to copy all the files starting with chap on the remote
machine warlock's /usr/bill/book directory (assuming you have
permission) to your own /usr/bigbook directory, you would issue the
command
uucp "warlock!/usr/bill/book/chap*" /usr/bigbook/
You can also specify hops in the machine transfers by adding the extra
machine names to the command. This requires permissions to be set on
all the machines that the hop will pass through, and is seldom done.
Although you probably don't want to do this either, you could transfer
files from one remote system to another by specifying their names on
the command line like this:
uucp arthur\!/usr/lib/uucppublic/bigfile warlock\!/usr/lib/uucppublic/
This sends the file from the arthur system to the warlock system. In
most cases, the users on either of the two remote systems would issue
the commands, relieving some of the file permission problems.
Checking on Transfers
You can check on the status of transfers that are scheduled but
haven't taken place yet using the uustat command. When you issue the
uustat command, all the UUCP transfers that are queued are listed. The
format of the list is:
jobID system user date command size
where jobID is the ID of the UUCP job, system is the name of the
system to transfer to (the first system in an address when multiple
hops are taking place), user is the user name who queued the job, date
is when the job was queued, command is the exact command to be
executed, and size is the size of the transfer in bytes.
If you issue the command as a user (not superuser), only your jobs are
listed. The superuser lists all jobs that are queued. If you are
logged in as a regular user and want to see all jobs, use the -a
option:
uustat -a
To cancel a queued job, use the -k option of the uustat command along
with the job ID. For example, to cancel job 17, issue the command:
uustat -k 17
You can only cancel your own jobs, unless you are logged in as
superuser, in which case you can cancel any jobs.
Summary
UUCP is quite easy to set up, as long as you follow the rules. Once
the configuration files are properly set, UUCP offers an easy way to
transfer e-mail, news, and files to other systems. UUCP is also one of
the easiest methods to establish a low-volume network because all you
need is a modem or a direct connection. Using UUCP to transfer mail
and files is as easy as using the usual mail and cp commands.
Although UUCP is less popular nowadays because of the LAN craze, it
does provide a simple, very low-cost network for those who need to
connect only a couple of machines. It's also great for connecting your
machine to those of your friends, allowing e-mail back and forth, and
really making your Linux system feel like a well-connected
workstation.
--
Enjoy Linux!
-----It's FREE!-----
※ 修改:．netiscpu 于 Jul 25 06:02:53 修改本文．[FROM: mtlab.hit.edu.cn]
※ 来源:．紫 丁 香 bbs.hit.edu.cn．[FROM: fengyun.hit.edu.]
Powered by KBS BBS 2.0 (http://dev.kcn.cn)
页面执行时间：615.128毫秒