Our setup:
Our desktop Linux machines are set up according to these instructions, which are summarized below:
- We start with a full install of RedHat 9.
- We make certain changes, such as replacing sendmail with nullmailer, configuring installed software, removing or disabling unnecessary things, and configuring a printer.
- We copy the /usr/local/ tree from one of our servers to the desktop machine. This gives the desktop machine all our locally installed system-wide software.
Software changes:
When we add or update software installed on our server, we can easily propagate those changes to each desktop machine.
As an alternative, software can also be installed inside a user's home directory. (No changes can be made to /usr/local/, since they would get blown away whenever we update, as explained in the previous paragraph.)
Before installing software inside a user's home directory, the following question should be asked: is this software something that other users might also want on their machines? If yes, we'd like to evaluate whether the software should be installed system-wide. In general, we prefer system-wide installation, unless: the software is exceedingly difficult to install or likely to conflict with other software.
Thus, installing software inside the user's home directory should be viewed as a last resort. But when it is appropriate, we can help you with this. The instructions are below.
Adding software in the user's home directory:
Note: re-read the above paragraphs---adding software specifically on one machine is considered a last resort. Please discuss your needs with us.
- Unless you know precisely what you're doing, you cannot drop in RPMs or other binary packages that are hard-coded to someone else's tree structure. Instead, you must do a somewhat custom install. These notes are guidelines, not step-by-step instructions.
- Inside your home directory, create a directory called "bin". Also create a place to hold your software's source and/or installation tree. For example, you might create a directory called "src" or "apps" or "software".
-
If the software is open-source, the build procedure is probably to run
some kind of "configure" script followed by various "make" scripts. The
standard GNU configure script takes a parameter to specify the final
install tree, which often defaults to /usr/local/. Thus you would do
something similar to the following (but never run these commands as root!):
./configure --prefix=/home/myaccountname/ make make install
-
If you simply copy a binary tree from CD or obtained over the web, then
you probably simply need to add links to your "bin" directory:
cd ~/bin ln -s ../apps/mysoftware/bin/foo foo
- Now you should be able to run the software. You might verify it is in your PATH. The command "which foo", where "foo" is the name of the command, should return "/home/myaccountname/bin/foo". If not, then you might check your ".bash_profile" or ".bashrc" or whatever to ensure that "~/bin" is part of your PATH.
