[Info-vax] A possible platform for VMS?

[Michael Moroney]

terry+googleblog at tmk.com writes:

>On the other hand, in modern Windows a good number of the patches can be
>applied without needing a reboot. Video drivers are an excellent example
>of wholesale driver replacement without needing a reboot. Quite a far cry
>from the old "Your mouse pointer has moved. Windows must restart to
>recognize these changes" model.

>Patch-in-place is becoming more popular. Recent Linux kernels now support
>live patching, for example. Perhaps this is something VSI can investigate
>as part of the port to a new architecture.

The vast majory of VMS patches to the base OS are really in the "You don't
need to reboot, but you won't see any of the effects of the patch until
you do" category.  That's because drivers and execlets are loaded into
memory (nonpaged pool) at boot time, and the files aren't touched
thereafter.  The OS won't notice or care if the file has been replaced,
deleted etc. until reboot.  There are exceptions, of course, such as
"Reboot NOW before something horrible happens/the bug we're fixing decides
to crash and reboot for you" or perhaps cluster instability if mixed
versions are in place, or when a driver is replaced along with a user mode
image where the user mode image depends on the driver change, and bad
things happen with the new image/old driver combo exists.  Also SDA of a
live system (ANALYZE/SYSTEM) can get confused when it tries to match new
driver files to an old driver in memory, but it will tell you that the
link dates don't match.

I'll suggest to VSI that a "Reboot at your convenience for changes to take
place" category be created for the "Reboot required" item (just yes/no 
right now).