Red Hat's CEO said the firm has put together a five-year strategic plan for the first time in its history, which will see it jack up its sales management as it looks to expand beyond its traditional markets. Jim Whitehurst is currently doing the rounds in Europe, as the firm gets ready to close on its fourth quarter and round …
Too much in the kernel already
I wish dists would stop heading off down the MS route and throw everything into the kernel that seems like a cool idea at the time. The kernel should be lean and mean. If you want KVM fine, have it as a loadable module. But do not integrate it completely into the kernel - it'll just be more bloat and more to go wrong.
Why distributions can't just stick to the stock kernel anyway beats me - its Unix in the 1980s all over again. The added value should be in the extras they provide in their distributions - they shouldn't mess around with the core software.
re: Too much in the kernel already
Don't think it is an issue to be honest. Redhat always provide seperate kernels for different roles. i.e.: RHEL5 uses kernel-xen or kernel-smp or whatever.
If you don't use it, then install a different kernel. Just like running a server in runlevel 3 if you don't use or need xorg/x11. Remove all services you don't need etc etc.
GNU/Linux lets you do this very easily which is miles away from windows architecture.
RE: re: Too much in the kernel already
Maybe that's exactly what is needed, a number of tightly integrated packages with as much in the kernal as possible so they can be run out-of-the-box for set application stacks. One of the reasons we buy hp-ux is they do the 11i v3 OS now in four bundles, which makes it easy to tie up all the patching, licences and support. It would be simple to create a "virtualisation bundle" for RHEL with KVM in the kernal, maybe a "cluster bundle", etc. Face it - making tight and lean kernal is fine for geeks, but lots of customers like the idea of just having something they can stick on a box and not have to fine tune. It would also allow differentiation in the pricing, so a cheaper support/licence bundle could be offered for a base bundle, then more expensive for one including clustering or KVM, etc. RHs big problem is not selling to the Linux fans, it's selling to Windows, Solaris, AIX and hp-ux users. Anything that makes that easier is a good thing, even if it upsets the fans.
Too much in the kernel?
I've yet to see a fully-functional NTFS module, and is almost 10 years due. What's the problem with adding stuff to the kernel? Even if the distros do put them all, most of them are kernel modules (which means they are not part of the kernel's main build) so it isn't like the kernel's stuck with a zillion features.
My only real complaint with RedHat is that they still can't provide mp3 support, and it's hard to get people to use Linux if they can't play mp3's out of the box.
ah the future
we will be back to chroots before 2010....