"screen" is a window manager that allows you to handle several independent
screens (UNIX ttys) on a single physical terminal; each screen has its own
set of processes connected to it (typically interactive shells). Each
virtual terminal created by "screen" emulates a DEC VT100 plus several ANSI
X3.64 and ISO 2022 functions (including DEC VT102 features such as line and
character deletion and insertion).
Since "screen" uses pseudo-ttys, the select system call, and UNIX-domain
sockets, it will not run under a system that does not include these
features of 4.2 and 4.3 BSD UNIX.
If you want to get a quick idea how "screen" works but don't want to read
the entire manual, do the following:
- call "screen" without arguments
- wait for the shell prompt; execute some commands
- type ^A ^C (Control-A followed by Control-C)
- wait for the shell prompt; do something in the new window
- type ^A ^A repeatedly to switch between the two windows
- terminate the first shell ("screen" switches to the other window)
- terminate the second shell
If you have got "vttest" (the VT100 test program from mod.sources) you
may want to run it from within "screen" to verify that it correctly
emulates a VT100 on your terminal (except for things your terminal is
incapable of handling, like 132 column mode and double width/height
By the way, "screen" can be used to compensate for certain bugs of "real"
VT100 terminals. For instance, our 4.2 BSD version of mille(6) garbles
the display on terminals of the VT100 family, but it works quite fine
when it is invoked from within "screen". In addition, "screen" enables
you to use EMACS on terminals that are unable to generate Control-S and
Control-Q from the keyboard or that require flow control using Control-S
and Control-Q. This is the reason why I have an alias like
alias emacs "screen emacs"
in my .cshrc file. Users who run EMACS on terminals which normally have no
trouble with Control-S and Control-Q will probably want to run screen with
the -fn or -fa options, or put the line "flow off" or "flow auto" in their
Before typing "make", look at the config directory and choose an appropriate
config file that you then copy to config.h and then edit it to
reflect the configuration parameters of your choice. Each define is well
commented to help you decide what to choose. Then give the Makefile the
To get screen to do everything it was designed to do, it should be installed
set-uid root. This enables it to modify the ownership of the pseudo-ttys,
update /etc/utmp, and read the current load average from /dev/kmem.
Failing to install it suid root (e.g. if you fear a trojan horse) doesn't
have any major disadvantages, except that w(1) and some other utilities
will display only "screen" as the current process, and the pseudo ttys
(ptys) used for the virtual terminals won't have their owners set. Screen
can provide you some pty security by opening the ptys exclusively, but
this has the unfortunate side-effect of keeping your own subprocesses from
being able to open /dev/tty.
It is also possible to create a special "screen" group and give the file
/etc/utmp screen-group write-access, and give the file /dev/kmem screen-group
read access to get all the features except the pty-ownership changes. Some
people also install it set-gid "kmem" or set-gid "utmp" to give it one or
Many people install screen as a normal program with no special privileges
and still find it to be indispensable. Give it a try.
Send your questions, comments and bug reports for screen to:
jnweiger@immd4.informatik.uni-erlangen.de
projects / unix-history / blob