In a message on 02-09-95, Stephen Lindholm said to Peter Hansen:

PH> It sounds like the 3-4 millisecond range is not quite right.

SL>Device drivers are never preempted. I can write one that occupies 5
SL>seconds between yields. It depends on what drivers you have loaded.

Obviously a device driver can do whatever it wants.  However, if one
follows the design guidelines one will _not_ disable interrupts for more
than 400 microseconds and will therefore not encounter a problem with 5
second interrupt latencies (which would kill the system anyway).  

And since we're talking (soft) real-time, the system designer would
obviously need to check that the device drivers not under his control
were not written by bumbling idiots.  I doubt very much that any
standard device drivers included with Warp, for example, disable
interupts for longer than about 75 microseconds at a time (max!) on a
486DX33.

Your point is theoretically correct, but moot.

------------------------------------------------------------------------
 Peter Hansen                                     Engenuity Corporation
 Internet: peter{at}engcorp.com                    Guelph, Ontario, Canada
___
 * MR/2 * Now cruising at Warp 3!

--- QScan v1.14b / 01-0348