🚀 go-pugleaf

re: Mike's comments on setting clocks
>
>    - Setting the clock forward is OK.
>    - Because timestamps in log files (both system and IMAGE, at the least)
>      contain the LOCAL time, resetting the clock can cause transactions to
>      be out-of-sync in the log file.  This is not good.
>    Solution:  Do not allow system clock to be reset while system is up.
>
> Now, the astute reader has no doubt realized that the solution does nothing
> to address the problem.  You shut down your system, reboot, reset the system
> clock back one hour, and your timestamps are out of sequence, exactly as if
> you had done this while the system was running.  It just requires more time
> and effort to shoot yourself in the foot.

The difference is that shutting the system down also closes the current
set of log files, and the reboot (on the new time) starts a fresh set,
hence there is no 'backwards' time jump within a single log file.  Now
the 'proper' system manager will of course backup his system before
shutting it down (since this is always an opportunity for some kind of
problem to rear it's head).  Hence, any recovery activities for future
failures will depend on the new logs only - and will not see the jump
backwards.

> In MPE/iX 5.0, a CI-level command or utility (I'm not sure which) is being
> provided to reset the system clock.  It works, I am told, by either speeding
> up or slowing down the system clock until it gets to the 'new' time -- thus,
> there is no 'jumping' of the time either forward or backward.

Also 'off the record', the 5.0 command as I've heard the story will slow
down or speed up the clock until the new time is reached.  There was also
some mention of a shoot-yourself-in-the-foot option that would still
allow the brave SM to demand an immediate clock change at-your-own-risk.
Again, this is just the 3rd hand rumors I've heard.

> Another piece of our MPE heritage gone forever.  Sigh.   :)

Personally, I think the advent of AIFs slowed down a lot of 'interesting'
hacking that went on :-(  The cost of progress?

dave@cup.hp.com

> NOT speaking for HP, officially or otherwise.

ditto