Microsoft exec says Namibia and Iran have recently changed DSTrules

[Jesper Norgaard Welen]

It is great to see Microsoft finally putting some muscle behind timezone
updates. In the spirit of getting it right, here are some comments on the
presumably latest update http://support.microsoft.com/kb/931836/en-us (note
the english version, because you will receive a different version according
to your IP, I received a spanish version here in Mexico).

(1) Central Brazilian time: There are two states on GMT-4 with DST which
would correspond to the timezone created; Mato Grosso and Mato Grosso do
Sul. Manaus is a city in the state of East Amazonas, which is on GMT-4
without DST and therefore does not apply as title. This also concerns the
timezone in the update, which should have a different title than Manaus. See
for instance decree 5,920 on
http://www.planalto.gov.br/ccivil_03/_Ato2004-2006/2006/Decreto/D5920.htm .

(2) Namibia: it seems to me that a wrong inversion of DST vs Standard time
is exactly what has been done *now*. Namibia is well below equator and
therefore will not benefit from having DST between April and September
(their winter time, just like Brazil or Australia or Chile). The start/end
times have been inversed both in the web page and in the update timezones.
It should start in September and end in April.

(3) Egypt: The DST given is from Last Thursday in April at 23:59:59.999 to
Last Thursday in September at 23:59:59.999. Actually the timezones for XP
only expresses 23:59:59 without the extra precision; I don't know if this is
a Windows Vista refinement. However, the tz database starts Last Friday in
April at 0:00. This might seem equivalent, but there is a week of difference
(instead of a 0.001 fraction of a second) in 2009, where the last Thursday
of April is 2009-04-30 while Last Friday is 2009-04-24. The End date is
correct - WTE has Last Thursday of September at 24:00, while tz database has
last Thursday of September at 23:00 (standard time), both which are not
compatible with Windows timezone definition, so the given approximation is
the closest possible.

(4) Beirut: I'm guessing that Lebanon is just using EU-like DST, while
applying it at midnight instead of at 3:00 and 4:00 as for instance nearby
Cyprus. Tz database has Lebanon DST to end Last Sunday in October at 0:00
similar to EU. Microsoft has DST ending in Last Saturday of October at
23:59:59.999, which is again 1 week wrong in 2009, where Last Saturday of
October is 2009-10-31 while Last Sunday of October is 2009-10-25.

(5) Jordan: Any of tz database and WTE and
http://www.timeanddate.com/worldclock/timezone.html?n=11 have DST to end
Last Friday of October just like 2006 (even in 2008 and 2009). In 2005
Jordan DST ended Last Friday of September. Do you have any authoritative
information to support this claim of Last Friday of September for 2007? If
not, you should change it to Last Friday of October instead of September.

(6) Santiago: The Microsoft timezone for Chile starts DST Second Saturday of
October at 23:59:59.999 and ends DST Second Saturday of March at
23:59:59.999. Note that this corresponds to First Sunday on or after 9.th at
0:00. Although there were several implementations of 23:59:59.999 with wrong
sideeffects above, this one is the only way to get it right. I thought I'd
just mention that because you are likely to begin wondering about *all*
implementations of 23:59:59.999.

Regards,
- Jesper Nørgaard Welen

>-----Original Message-----
>From: John Gray [mailto:johngray at winse.microsoft.com] 
>Sent: Viernes, 02 de Marzo de 2007 14:43
>To: tz at lecserver.nci.nih.gov; tz at lecserver.nci.nih.gov
>Subject: RE: Microsoft exec says Namibia and Iran have recently changed
DSTrules
>
>It was somehow garbled.
>
>Namibia: We had an inversion of DST vs Standard time in the original
KB938388 due to some confusing interpretations. The clock change dates were
correct, but this fixed the DST period.
>
>Iran: We didn't get official confirmation in time for Vista or the Nov
KB938388 on their elimination of DST.
>
>The only two recent changes that affected Windows were Newfoundland and W.
Australia, that were changed after our KB938388.  We aren't granular enough
to track things like Nunavut (not Nunavit) yet. :)
>
>From the official KB at http://support.microsoft.com/kb/931836  ...