From: "Frank Haber" 

>an event log file records each timestamp in local time

Thanks, John.  As you sussed, I was fishing, genuinely not knowing what the
"temporal philosophy" of Windows is.  Geo told me what I already
knew, so let me expand what I'm not sure of.  Not being any kind of o-o
programmer, I imagine a stack like this:

1. At root, the hardware just "counts Julian," incrementing an
arbitrary integer value, to as fine a granularity as can be managed.  The
opsys has a synched clock that is periodically updated from the hardware. 
It also queries time servers, and corrects the hardware count periodically.
 From now on, the software count is asynchronous with the hardware and the
world, because we do NOT have a realtime design here."

2. The opsys has tables for Zulu offset, and a table marked "stupid
and illusory" for the DST rules (g).

3. The opsys's APIs use any algorithm that's convenient for calculations, and......

(Frank's knowledge falls off a cliff here)

4. Emit/offer ("expose" is the current jargon?) time in ?Zulu
(?local) for apps to do their stuff and hand the results back to (? a stack
of stuff that crosses ring 0 and manages the way objects pluck the results
they need.)  This is (?separate from the|?the same) system that manages
time-critical stuff like interrupts, process priority, user-experience
visual refreshes, and fairness?

5. Final results are thrown back into the above API maelstrom for conversion
?back to local and, optionally, display to the user.  All communication via
LAN or WAN must include a GMT time and offset, and something to indicate
whether the value is ca. Bush2 or Clinton, etc.  This has gotta be
cross-platform, right?  And there's gotta be an RFC?

(I know that fully explaining #4 would require a book or ten to explain
properly.  Give me an exec^H^H idiot's paragraph. And big thanks.)

--- BBBS/NT v4.01 Flag-5