Pádraig Brady <address@hidden> wrote:
> I argued to hide all that in the kernel 3 years ago
> so a seperate interface was not needed:
>http://lkml.org/lkml/2003/11/10/64
I agree.
It would certainly be better from a user-interface standpoint
to be able to tell the system that a process has priority X,
and the system would just do the right thing, wrt both CPU and I/O.
So I guess now I'm saying that this functionality belongs in a
program like `nice' that does both things. Having to resort to
using TSTP and CONT signals is a kludge, but maybe necessary to
provide this functionality even on systems without adequate kernel support.