Re: [patch] lockless, scalable get_pid(), for_each_process() elimination, 2.5.35-BK

[Cort Dougan]

I'm also a big fan of "Do that crap outside the kernel until it works
properly" type projects.  If you want to talk about real-time scheduling as
an example... looks at the trouble it's causing now.  I think it makes my
point for me very strongly.

The Linux view should not be that N-way boxes are its manifest destiny.
The same goes for thousands of threads.  Linux works pretty well on 95% of
the boxes that it is being run on.  Lets not screw that up to fix the other
5%.  Try some fixes _outside_ the main kernel for a while, find a workable
solution and then merge it in.

The Linux bus is getting really top-heavy because of some macho-features.

} Like real time scheduling or embedded systems for example?
} Be careful what you ask for ....