At 9:11 PM -0800 3/15/00, Todd Whitesel wrote:
>> Correct. 68ks have drastic hardware interrupt problems which Apple
>> ignores but (and I forget the reasons why) we can't. (Oh, wait, it's
>> that MacOS doesn't care if everything stops for a disk spinup...
>> right.)

I think MacOS reads/resets the OS version of time after every floppy
operation, or other action that might turn off interrupts for a length of
time.

>Both BSD and MacOS have 'tick counters', init'ed when the system boots up.
>
>The problem is, BSD insists on trusting the tick counter for everything,
>but on MacOS apps are told not to trust it much at all.
>
>On MacOS if you ask for the date/time, the toolbox goes straight to the
>clock chip and ignores the tick counter.
>
>Why don't we just write a user-readable clock chip device, and hack xntp
>to use it as an external source? That'd fix this.

Because the clock chip is only readable to the 1-second level, I believe.
It keeps time internally as you would expect, but the interface is not what
you need.  Also different models have different degrees of clock fidelity.
ntpd does a pretty comprehensive job of "fixing" things, but it needs to be
able to access all the timing variables in a way and to a level of
precision that many (most?) clock hardware implementations wouldn't support
even off the Mac-68k.

This thread has gotten a bit off-topic.  I've cc-ed port-mac68k, and any
future discussion should probably go there exclusively.

Signature failed Preliminary Design Review.
Feasibility of a new signature is currently being evaluated.
h.b.hotz@jpl.nasa.gov, or hbhotz@oxy.edu