Hi Tim, Brooks, James, Robert, and Paul,

I appreciate everyone's thorough follow-up. This discussion was extremely
helpful in framing the issue correctly.

My take-away is that the "First Thursday" behavior in the Windows registry
is an expected, if legacy, implementation detail—effectively a placeholder
to bridge the data structure requirements rather than an error in the
underlying tzdb. Since 2027 is modeled as a clean transition to permanent
UTC-5, there is no risk of the 2026 DST rules affecting the new year.

I’m satisfied that there is no issue with the Manitoba Standard Time update
itself. Thank you all for helping me verify that the source data is
functioning as intended.

Best regards,

David Abbott

On Sat, Oct 17, 2026, 9:41 AM David <[email protected]> wrote:

> Hello tz maintainers,
>
> I’m running into an issue with Microsoft Windows time zone data and hoping
> you can confirm whether this stems from tzdb itself or from Microsoft’s
> conversion.
>
> Issue summary. For the end of 2027, Windows appears to calculate the DST
> end date as Thursday, January 7, 2027, instead of Friday, January 1, 2027.
> This one week shift affects local time conversions around New Year’s for
> Manitoba Standard Time in Microsoft’s update.
>
> Specific details. IANA zone ID, likely America/Winnipeg. Date and time in
> question, end of 2027 DST transition, expected Friday January 1, 2027.
> Expected offset, per tz rules for that zone on 2027-01-01. Observed offset
> in Windows, appears to use 2027-01-07 instead. Windows version and
> environment, Windows 11 and UWP classes, which pull from tzdb via
> Microsoft’s registry mapping.
>
> Background: Windows 8.1 and later derive tz data from tzdb but convert it
> into the Windows registry format. I’ve been working with Microsoft Support
> but have been referred between engineers without anyone reviewing the
> actual historical offset rules.
>
> Could you confirm whether the DST end for 2027 in tzdb is expected to be
> Friday January 1, 2027, and not Thursday January 7, 2027? If tzdb is
> correct, that would point to an issue in Microsoft’s mapping or handling of
> the end of year transition for Manitoba Standard Time.
>
> Thanks for maintaining this critical dataset.
>
> Best,
>
> David Abbott
>

Reply via email to