The patch screen-3.9.4-notmp.patch that the screen-3.9.5-4 rpm was built with caused a lot of network traffic between two fast computers on our network. A user ran screen on a linux computer with RedHat 6.2 and his home directory was NFS mounted from a SunOS 5.6 computer. As SOCKDIR was on an NFS-mounted directory this caused problems. About 5000 packets per second of NFS traffic. The man page for screen said: CUSTOMIZATION The "socket directory" defaults either to $HOME/.screen or simply to /tmp/screens or preferably to /usr/local/screens chosen at compile-time. If screen is installed setuid- root, then the administrator should compile screen with an adequate (not NFS mounted) socket directory. If screen is not running setuid-root, the user can specify any mode 700 directory in the environment variable $SCREENDIR. Given that a lot of people have NFS-mounted home directories I consider the screen-3.9.4-notmp.patch patch file to be perhaps causing more problems than it solves. I cannot see the reason for it being there in the first place? But I would like to know. Removing this patch from the source RPM distribution and rebuilding the package solved our problem. Eskil... :-)
It's *much* simpler from a security standpoint; it avoids lots of wrangling to allow both root and a user to use screen at the same time, and it avoids making various world-writable directories and files in /tmp.