On Tue, 27 Oct 2026 at 01:47, Paul Eggert via tz <[email protected]> wrote:
> The first one implements your suggestions. The second one rewords our data > to better match the wording of older Brazilian decrees (unlike your > suggestions, this doesn't change any timestamps). Thanks; I'd come to broadly similar conclusions. However, it's worth noting that phrases like "até 31 de março", even when defined in law to "include" the last day, still carry some ambiguity — is that last "legal day" that is considered the *daylight time* day, or the *standard* day? That is, does the "last day" officially end at 24:00 Daylight Time or 24:00 Standard Time? Different countries seem to have taken different approaches over the years, but 24:00s would not have been an uncommon interpretation in the first two-thirds of the 20th century. It's entirely possible, then, that the explicit mention of a few 01:00 fall back transitions such as 1950 and 1966 could be an indication that many of the other implied midnight "fall back" transitions should be interpreted as 24:00s instead of 24:00, with clocks effectively falling back from 25:00 to 24:00 (that is, 01:00 to 00:00 the following day). So I turned to contemporaneous newspaper reports, which unfortunately didn't clear things up. While both the 1950-04-13 decree and the 1953-02-24 decree use the same implicit phrasing without naming a specific hour for the transition, *Jornal do Brasil* reported the similarly-phrased changes somewhat inconsistently: The 1951-03-31 edition describes a 01:00 transition: "a partir da uma hora da madrugada (Hora de verão) do dia 1.º abril proximo vindouro" "starting at 1:00 a.m. (daylight saving time) on this coming April 1" https://news.google.com/newspapers?id=VpxTAAAAIBAJ&sjid=44YDAAAAIBAJ&pg=6869%2C3826611 The 1953-02-28 edition describes a 00:00 transition: "a partir de zero hora de amanhã" "starting at midnight tonight (going into tomorrow)" https://news.google.com/newspapers?id=NGkPAAAAIBAJ&sjid=JocDAAAAIBAJ&pg=1610%2C3886448 Other years were either less clear or harder to find. Given this inconsistency in reporting, it's possible the exact transition times were not observed terribly strictly, so I'm not inclined to change things further without clearer contemporaneous reporting of what folks actually ended up *doing*, regardless of what the decrees or the newspapers may have told them to do. That said, we should at least make note of where these 24:00 transitions might mean 24:00s, so I've added commentary to that effect in the attached patch further to Paul's. In looking through the decrees myself, I also found another explicit 24:00 in a nearby spring forward and changed that as well to match. (Our old 0:00 timestamps in the AT field here date back to earlier zic that couldn't handle 24:00 properly.) Additionally, I noticed that the situation around scheduled DST end dates in 1963–1964 was a little confusing, as each successive decree superseded the end date scheduled in the last, so I updated our comments there a bit to remove an incorrect implication it seems we'd made. -- Tim Parenti
0001-Additional-commentary-for-Brazil-1963-1965.patch
Description: Binary data
