You see, it has been a good way to understand a little better the tz format. And a nice patience exercise also :) Paul Eggert wrote:
Thanks for your proofreading of the history of time zones and daylight saving time of Brazil. Here are some thoughts about your comments.
From: Rodrigo Severo [mailto:rodrigo.lists@fabricadeideias.com] Sent: Thursday, October 16, 2003 7:31 PM
1. The effects of decrees 52700 (1963-10-18) and 53071 (1963-12-03) weren't correctly represented by the data as what they really have done was to make a set of states start DST at 1963-10-23 00:00 and the rest of the country to start DST at 1963-12-09 00:00.
As far as I can see, the current tz data does that. The tz source looks different from what you're proposing, but the actual transitions already equal what would be generated by the proposed change. For example, for America/Manuas the 1963 transition is currently this:
Mon Dec 9 03:59:59 1963 UTC = Sun Dec 8 23:59:59 1963 AMT isdst=0 Mon Dec 9 04:00:00 1963 UTC = Mon Dec 9 01:00:00 1963 AMST isdst=1
and this transition isn't affected by the proposed change. So I don't see the need for this part of the change.
You are right. I just used zdump for final check of my review and I was still grasping the tz data format. Just checking: the only real difference between the present data and my proposed patch in this matter would be the creation of a new timezone (America/Brasilia). As this new timezone is only necessary to reflect a time difference between America/Sao_Paulo and America/Brasilia BEFORE 1970 it shall not be created.
2. The starting time of DST for 1966 was wrong as the 57843 decree stablishes that DST shall [end] at 1966-03-01 01:00 and on 1st of March 00:00 from 1967 and on.
I'm inclined to think that this may be a typographical error or ambiguity in decree 57,843 <http://pcdsh01.on.br/HV57843.htm>. It'd be unusual for Brazil to change at any time other than midnight. This reminds me of the 24-hour error in Decree 4,844 this year (though the latter error was corrected after a couple of days). However, I don't read Portuguese so I'm not really qualified to interpret any wording problems in decree 57,843.
I would not call this error typographical, but probably something worst: the guy that originally wrote the decree had this fancy idea of ending DST time in a "better" hour or something like that. Anyway I agree that the brazilian tradition of time changes at zero hour and the tiny impact this change would cause calls for something simple as the comment you proposed below, not a new rule as I proposed.
5. America/Sao_Paulo zone had a DST during the whole 1964. I can't see where it came from. There is no decree mentioning it.
It's not all of 1964, it's only the period from 1963-10-23 00:00 to 1964-03-01 00:00. I think issue (5) is the same as issue (1) above. It's merely a different way to split the work between the Zone and the Rule lines; the actual transitions are the same regardless of whether the proposed patch is applied.
Yeah, I understand that now.
Here's a diff to reflect the above comments. I'll include something like this in my next proposed patch.
I believe your patch is fine (much better than mine BTW). Rodrigo Severo