"Self support" is the support mode for just about every Linux distribution unless you choose one that also sells support contracts.

Red Hat points RHEL at the server/enterprise market. After trying to sell shrinkwrapped retail Linux to home users, Red Hat wisely realized there's no money in that.

Red Hat is an adamant fOSS/open source supporter. CentOS and Scientific Linux, then, are recompilations of RHEL with the Red Hat logos and other trademarks removed.

Like RHEL, CentOS targets servers but it works just fine on the desktop. For its intended market, Red Hat has no reason to try to stay current re: desktops. But, with a bit of research and effort, you can build a very nice Gnome 2 desktop with CentOS (I've done it). To do that, you will very likely want to pull packages from independent repositories. You need to be careful about that because it is quite possible to overwrite critical CentOS packages with stuff from one of those indy repos. (This can happen with *any* distribution when you install software from other sources, including compiling it yourself.) There are easy ways to avoid this and they are explained on the CentOS wiki.

The CentOS forums are active and knowledgeable. You won't find much desktop talk there but it's a friendly place. Do read the docs, etc, though, before posting.

CentOS is pretty fast pushing out bug fixes from upstream, i.e., Red Hat. Usually in a day or so.

In my experience, CentOS is very reliable. The kernel, while numbered in the 2.6 range, contains much that Red Hat has backported from later and current kernels. (They list all those somewhere on their site.)


The RHEL and, hence, CentOS, online docs are excellent.

The downside is that apps on CentOS are often an iteration or two behind the curve, so it is no place to be if you insist on the latest versions of everything. (Like current versions of Gnome or KDE.)

Fedora is sponsored by Red Hat as something of a proving ground for code that may, or may not, make its way into future RHEL releases. As such, it's a pretty bleeding edge piece of work, much different from CentOS. Go there if you want the latest. Fedora does a release every 6 months and there are often significant changes from one release to the next,

I use Centos 6 with Fluxbox, Thunar and Rxvt as my base GUI. As you said it is easy also to use Gnome etc if you wish. It is true that the kernels are still at 2.6 but with care and the kernel headers & compile tools correctly installed you can build many of the later apps including some but not all 'bleeding edge' ones if you require them. Because of its close relation to RHEL it is ideal for students and developers alike although the latter will usually have their own preferences. In the main Centos is a very stable start point if you want to build your own system and they offer the install media to do this easily with a netinstall cd which will install just a server if you so require, to which you can add other services as necessary very easily. The procedure is as follows.

1. Get the netinstall cd
2. Install base system and set up Epel repo, kernels and compile tools
3. Build whatever you like on it.

I don't think it can get much better than that !

I use the El Repo site to install an Nvidia driver, add EPEL, Google (for Chrome), and then the Nux repo to grab its codecs. My HP printer needs an HP driver so I grab those tools from the Linuxtech repos. Once set up, I disable the last two repos. The Remi repos always offer the current Firefox and Thunderbird.

Netinstalls are very nice. Fedora 18 still has work to do on its new installer, but the netinstall in the beta is very good, with a screenful of options.

@joncr You can install the GENUINE Nvidia drivers (as opposed to the kmod ones) and the open source Chromium browser without the so called spyware if you go to the links there. They are for Centos 6 but both should be fine or pretty much any distro.

The problem with using Nvidia's installation package is that you need to repeat the installation each time the kernel is updated. That's easy to miss when you boot directly into X. If you do miss it, your next reboot will fail. The kmod package has the same drivers and handles the recompile automatically behind the scenes when a kernel update happens.

I've used Nvidia's ".run" install on Slackware because there I don't boot into X. If I update the kernel, it's simple to redo the install.

At kernel update you have a choice. Leave as it is, OR update and then re-do the kernel headers/kernel-devel + Nvidia drivers. The 2nd choice takes about 3 minutes for everything, so it's not a huge problem. Genuine Nvidia drivers are way better than the kmod ones. Thats the way I prefer and I use just the basic X with GDM and Fluxbox, but it doesn't really matter what you use.

The kmod package is simply a wrapper around the "Genuine" nvidia driver.

OK, just to save manually updating kernels etc. Not much in it I suppose, thanks for pointing that out.