Philip, my comments below
Philip Warner [pjw@rhyme.com.au] wrote: I'd like to clarify which year should be compared in from/to fields for rules (and zone lines). It seems reasonable that the AT field determines which year is to be compared (wallclock, utc or standard) -- is that correct?
This is really quite simple because we don't allow AT field -1:00 (the day before) or even 24:00 (because some C compilers won't support it) so the year is always what the rule specifies, in the below 2 rules 1999 and 2000. However, if using u or s then effectively we could be talking about *local* time of another day, even another year. Yes, the AT field determines which year, that's correct.
To give specific example:
Rule Test 1999 only - Jan 1 0:00 2:00 - Rule Test 2000 only - Jan 1 0:00 0:00 -
These are both wallclock rules, and suppose they are used in a region that is GMT+5. Then at UTC 31-Dec-1999 17:00 the WC time will turn to midnight....is that the point at which the clocks would be adjusted?
This is a bit weird, you are specifying in the first rule that at 0:00 1-Jan-1999 (local time) you are setting the clock 2 hours ahead, e.g. it will be GMT+7, starting year 1999 effectively at 2 o'clock ... This would be at UTC 31-Dec-1998 19:00 (5 hours before) ... when local time effectively changes from midnight to 2 in the morning. The next rule states that at UTC 31-Dec-1999 17:00 (local midnight) the clock is set 2 hours back again. So your interpretation are not correct, since wall clock will not turn to midnight, it will rather leave it! Note that in the example you would celebrate New Year twice from 1999 to 2000 in your local place! So in effect, you might mean the right thing although I don't like the formulation.
And if so, in the following example:
Rule Test 1999 only - Jan 1 0:00 2:00u - Rule Test 2000 only - Jan 1 0:00 0:00u -
would the clocks adjust at UTC time 1-Jan-2000?
Here the AT field is 0:00 but the SAVE field is 2:00u - this is not possible as the u or s postfixes belong in the AT field. Suppose you had written the rules like this: Rule Test 1999 only - Jan 1 0:00u 2:00 - Rule Test 2000 only - Jan 1 0:00u 0:00 - Then if we presume GMT+5 then effectively the change takes place UTC 0:00 1-Jan-1999, e.g. at 5:00 1-Jan-1999 local time, where the clock is set forward 2 hours. Then at UTC 0:00 1-Jan-2000, the clock is set backward 2 hours, from local time 7:00 1-Jan-2000 to 5:00 1-Jan-2000. Note that in all cases I presume that there were no previous adjustments to local time, which would alter the calculations. I hope this clarifies the use for you ...