On 7/7/2014 3:53 PM, Robert A. Rosenberg wrote:
OOPS. I might have it backwards but my principle still stands. There
only ambiguous time stamp is the switch day when the 1AM hour can
occur twice. On that day when you see get to 1:59AM and roll back to
1:00AM again (and thus see timestamps in the 1:xxAM hour earlier than
a previous timestamp) you know to adjust the UTC offset. When you skip
the 2:xxAM hour then you know to use the correct offset for 12M to
1:59AM and to switch to the other offset when you see 3:00AM.
For data produced near the "fall back" event:
If the input data is not ordered chronologically (if a 0415 datum
can follow a 0430 datum), then a 0115 datum following a 0130 datum tells
you nothing about which offset to use for either.
If the data is not reasonably dense, then a 0130 datum following a
0030 datum tells you nothing about which offset to use for the former.