But if the 960 is simply running the i2o or similar i/o subsystem,
the latency between the interrupt assertion and a useful response
may increase greatly --- not to mention that the 960 will be running
a MS written mini-os that may not be exceptionally speedy.
Since it is obvious that MS wants to move into RT control, I wonder
how their RT strategy will deal with these "intelligent" i/o subsystems.
In particular, they have a source license to Radisys to do something that
seems similar to RT-Linux using IRMX as the rt kernel (the
info releases from Radisys are vague on what they are exactly doing).