Yes. This is the approach the FSM has taken. I once spoke to one of 
their developers on a trip to Munich and he was convinced that this was 
the right way to do hard real time with a Linux kernel. If you really 
narrow your opinion of what an RTOS should be, his arguments made some 
sense. But, in the world of deterministic, hard real time, this 
solution, IMHO, falls flat. I don't believe that there is a 'right' way 
of doing hard real time in Linux, because I don't believe that the way 
exists at all.

David Rhodus wrote:
> On 10/26/05, Jim DeLisle <jdelisle@swellsoftware.com> wrote:
> 
>>'Real Time' does not refer to how fast an OS runs. It refers to
>>determinism. In other words, the latency between the IRQ and the ISR
>>vector is consistent under any circumstances. A 'soft' real time OS,
>>like WinCE, for example, does not provide this level of certainty. One
>>version of real time Linux actually runs a small real time executive
>>which, in turn, runs Linux as a task. The real time elements are not
>>built into the Linux kernel itself, and applications that wish to
>>participate in real time scheduling do so through an interface to the
>>executive.
> 
> 
> Does anyone know if this is the same method FSMLabs uses for their RT BSD's ?
> 
> --
>                                             -David
>                                             Steven David Rhodus
>                                             <drhodus@machdep.com>
> 

-- 
Swell Software, Inc.
jdelisle@swellsoftware.com