Sorry if I sounded dismissive; I had assumed you were using Method 1 from
this official HOWTO. I'll explain exactly what I've been talking about.
All Unix-like operating systems have the concept of a Superuser, which is normally called \"root\". It is identical to the SYSTEM account in Windows; it can do anything it wants to anything on the system, regardless of permissions. For example, only root can write directly to video memory, and root can read, write, create, and delete files no matter who owns them and/or their permissions.
\"su\" stands for switch user. It allows you to change what's known as your effective UID -- the process you launch functions under the user you choose. You must provide the password for the user you wish to become, if a password is set. Running \"su\" without an argument means \"switch to root (UID 0) and run a shell\"; there are a number of arguments that can change exactly what su does (the most common being a different UID or username to switch to). root can su to any other user without providing a password.
Ubuntu locks (disables) the root account by making its password un-enterable. Passwords, like all user account information, are stored in encrypted form in /etc/passwd. By making the encrypted password \"x\", no password you can enter can possibly match it. Thus, you can't su to the root account, but it still exists and can be used (for example, C programs can call setuid(0) to try to become root).
So then how do you do root-only stuff, like starting/stopping daemons or editing certain configuration files? Ubuntu comes with sudo, which means what it says \"switch user and do\". It's a beefed-up version of su, allowing you to become root without having to provide the root password, just your own. It works by maintaining a list of users who are allowed to become root in /etc/sudoers, and if a user in that list runs sudo, prompts them for their password. Providing the correct password lets the user become root and do whatever they told sudo to do. \"sudo kate\", for example, will prompt for your password and then launch the Kate text editor as root, letting you edit any file on the system.
This is a better practice than sitting at a root terminal. I can't count the times I've done stupid things, forgetting I had run su, and hosed parts of or all of my system. sudo reminds you you're about to do root stuff by prompting you for your password, and doesn't leave you in a state where you can continue doing potentially dangerous stuff. But it does get tedious if you need to run a lot of commands as root, so \"sudo -s\" does essentially what su does -- drop you at a root shell -- but, again, using your own password instead of root's.
Anyway, \"apt-get\" is the Debian package manager (*buntu is a Debian derivative). \"apt-get install x\" will download and install package x. The Ubuntu peeps are nice enough to provide a stable NVIDIA driver package called nvidia-glx in the repository. So \"apt-get install nvidia-glx\" will install that package. You have to be root to install software, so that's why the full command is \"sudo apt-get install nvidia-glx\". An alternative is to use Synaptic (or its KDE equivalent, which I think is called Kynaptic) to find and install the nvidia-glx package. The package includes a bunch of commands beginning with nvidia-, the three most important being nvidia-xconfig (which, when run as root, updates your /etc/xorg.conf file to use the NVIDIA driver), nvidida-glx-config (which, when run as root, enables or disables GLX, an extension which makes it possible to use OpenGL inside an X session, based on a command line argument), and nvidia-settings (the Linux version of the NVIDIA control panel applet).
I'm playing with the idea of messing with Linux myself, but that's a diffrent thread...
