Exercise 2 solution

    from datetime import datetime, timezone
    from zoneinfo import ZoneInfo

    def hours_between(start, end):
        naive = (end - start).total_seconds() / 3600
        real = (end.astimezone(timezone.utc) - start.astimezone(timezone.utc)).total_seconds() / 3600
        return naive, real

    ny = ZoneInfo("America/New_York")
    lon = ZoneInfo("Europe/London")
    shifts = [
        ("New York, 8 Mar", datetime(2026, 3, 8, 0, 0, tzinfo=ny), datetime(2026, 3, 8, 8, 0, tzinfo=ny)),
        ("New York, 1 Nov", datetime(2026, 11, 1, 0, 0, tzinfo=ny), datetime(2026, 11, 1, 8, 0, tzinfo=ny)),
        ("London, 29 Mar", datetime(2026, 3, 29, 0, 0, tzinfo=lon), datetime(2026, 3, 29, 8, 0, tzinfo=lon)),
        ("London, 25 Oct", datetime(2026, 10, 25, 0, 0, tzinfo=lon), datetime(2026, 10, 25, 8, 0, tzinfo=lon)),
        ("New York, 15 Jul", datetime(2026, 7, 15, 0, 0, tzinfo=ny), datetime(2026, 7, 15, 8, 0, tzinfo=ny)),
    ]
    for label, a, b in shifts:
        naive, real = hours_between(a, b)
        print(f"{label:<17} wall-clock {naive:.1f} h   real {real:.1f} h")

Output (run, and exactly as shown):

    New York, 8 Mar   wall-clock 8.0 h   real 7.0 h
    New York, 1 Nov   wall-clock 8.0 h   real 9.0 h
    London, 29 Mar    wall-clock 8.0 h   real 7.0 h
    London, 25 Oct    wall-clock 8.0 h   real 9.0 h
    New York, 15 Jul  wall-clock 8.0 h   real 8.0 h

Explanation: when both datetimes carry the SAME ZoneInfo object, Python
subtracts their wall-clock readings and ignores the offsets, so the answer is
always 8.0. Converting both to UTC first gives the elapsed time that really
passed: 7 hours when an hour is skipped, 9 when an hour is repeated, 8 on an
ordinary day. A shift paid by the hour needs the second figure.