Stupid Mandrake 8.1 question

[Coldfirex]

Why is it when I use the system update program in Mandrake it recommends not use it for new kernels? Why would it show it as new software and offer to install it if they dont "like it."?
Thanks

[Sneedes]

That doesn't make much sense does it. If they don't 'recommend' it, then why display the option to.

My only guess why they don't recommend kernel updates is because it may hose your system. A kernel upgrade can be a significant change to the system. And the point of Mandrake's update utility is to keep it simple.

Sneedes

[Forge]

Sneedes has it. It's a pretty major thing to swap kernels, and although it should work fine, it's not something you want to be doing if you don't need to.

Compiling kernels is great once you know your hardware, though, as it gives you a core to your OS that is perfectly tuned to the exact hardware you have.

[cRock]

Riddle me this:

On a FreeBSD system when you upgrade to a newer kernel you also upgrade to new userland utilities. This would tend to make sense because kernel changes can effect userland binaries that access low level kernel functions. To make life easy, you just run make buildword and it builds the entire base system, viola.

On Linux, it seems like the kernel is usually the only thing that ever gets updated. Does Linus just require that no major changes can occur that effect userland binaries (with the exception of version changes like 2.2 to 2.4)? Is there some other method to what appears to be madness? Does this partially explain why Linux systems don't always display that famed stability?

[SecretSquirrel]

On 2002-01-04 10:55, cRock wrote:
Riddle me this:

On a FreeBSD system when you upgrade to a newer kernel you also upgrade to new userland utilities. This would tend to make sense because kernel changes can effect userland binaries that access low level kernel functions. To make life easy, you just run make buildword and it builds the entire base system, viola.

On Linux, it seems like the kernel is usually the only thing that ever gets updated. Does Linus just require that no major changes can occur that effect userland binaries (with the exception of version changes like 2.2 to 2.4)? Is there some other method to what appears to be madness? Does this partially explain why Linux systems don't always display that famed stability?

I would expect that your latter guess may be right. Though I must say, taken a machine from early 2.2.x up to 2.4.1 with no difficulty. The only thing that complaing was the nfsd init script because lockd is no longer a seperate daemon (or something like that, I forget right off hand). Most of the userland utilities, at least that I use, are far enough removed from the kernel that it doesn't matter. One exception to this, for me, is IPtables. It's user utilities directly modified the behavior of kernel activites.