In message <20020709075832.A3176@ieee.org>Gregory McGarry writes
>cgd@broadcom.com wrote:

[...]

>> > RAS_CONTROL_PURGE_ALL_AND_INSTALL_ONE

>> (Why is the latter useful enough to merit being implemented?)
>
>As hard as I try, I can't think of any reason.

Atomicity of the purge/replace op itself?  

or.... If Mach mapped these onto Mach message exchanges, and then
replacing all current restartable-atomic-ops with a known, single op
(exec()-like?)  was dynamically common, then I could see it making
sense.