I really can't answer any questions about how the servicing works
presently.  I just checked my fully updated Windows 11, 25H2, 26200.9711
machine (I haven't updated to 26H2 yet though) - and even though I have the
new registry entries, and C:\Windows\Globalization\Time
Zone\timezoneMapping.xml has them too, it has an outdated ICU 72.1 and an
ICU zoneinfo64.res file corresponding to 2022g.  I also tested with both
UWP/WinRT and .NET APIs (at least the ones that use IANA time zones), and
everything is using old outdated data, presumably coming from this ICU
source.  I guess Windows has been overlooking these in servicing updates.
I'm really not sure why.

As for when this was introduced, search for Microsoft support articles
describing the Windows NLS to ICU transition - there are several.  There
are also some that describe when .NET 5 first started using ICU, and you'll
find one with my name on it around the .NET 6 announcement.

But yeah - sucks this hasn't been kept current.  My only advice for Windows
developers then would be to either use app-bundled ICU so you can control
the data yourself, or for .NET consider using NodaTime (a JodaTime-like
third-party library, with its own tzdata embedded).

I'll see if things are any better in 26H2 after I upgrade.

-Matt


On Wed, Oct 21, 2026, 12:29 PM Tim Parenti <[email protected]> wrote:

>
> On Mon, 19 Oct 2026 at 23:53, Matt Johnson-Pint <[email protected]>
> wrote:
>
>> Note that these posts describe Windows time zones in terms of their
>> "legacy" or "registry" time zone identifiers
>>
>
> Thanks for the detailed explanation, Matt, and for helping us speak more
> clearly on this and where it fits into the overall ecosystem.  And thank
> you also for your years of service in those efforts!
>
>
>> TLDR: Yes, Windows uses IANA TZ data (via ICU), and it ALSO maintains its
>> legacy registry format because older APIs require it.
>>
>
> It's helpful to understand that most modern things in Windows are able to
> make use of IANA tz data.  I hope you can help me (and presumably others) to 
> understand
> the actual dependencies, then, to the extent you are able to characterize
> them.
>
> I see ICU releases through 78.3 which only goes up to tz 2026a.  More recent
> patch releases of CLDR <https://cldr.unicode.org/downloads/cldr-48> have
> followed closely behind our more recent tz releases — generally within
> hours and only updating tzdata — and the latest CLDR 48.2.3 has the latest
> tzdata 2026e.  So when you say Windows/Microsoft are using tz data from
> ICU, does this include these other CLDR releases?
>
> And then, I presume, in much the same way as other vendors like Apple and
> Android, someone at Microsoft begins the internal processes of testing,
> validating, and deciding when to package these updates for downstream
> release, which ultimately results in the types of announcements we see on
> the DST Blog?
>
> Assuming the typical case that existing zones' definitions are being
> updated, with nothing out of the ordinary like a new zone or multiple
> transitions in a year, is it safe to assume that the data transformations
> into the legacy registry-based formats are reasonably automated?  Obviously
> there'd be a fair bit more editorializing than in an case that uses our
> data more directly, but it's not like anyone's looking at fresh tzdata and
> writing legacy registry entries by hand in the vast majority of cases
> anymore, right?
>
> In looking up information on CLDR's windowsZones.xml file to drill down
> into the earlier assertion that Windows uses it, I found a 2023 comment
> from you
> <https://github.com/tc39/proposal-canonical-tz/issues/16#issuecomment-1552142655>,
> which said "Windows has it's own copy of the mappings at
> 'C:\Windows\Globalization\Time Zone\timezoneMapping.xml'. It's in a
> slightly different format, but it is ultimately derived from the same CLDR
> source."  So presumably that's coordinated largely by Microsoft just
> controlling when both CLDR and tzdata updates roll out?
>
> Lastly: UWP was introduced with Windows 10 in 2015, and .NET 6 came out in
> 2021.  Would it be reasonable to characterize those dates, for those
> concerned, as roughly when these shifts toward IANA identifiers in Windows
> began (notwithstanding legacy registry-based implementations)?
>
> Thanks for any insights you can provide and to your contributions.  It's
> genuinely good to hear that there have been significant strides toward
> unified approaches where they are possible.
>
> --
> Tim Parenti
>

Reply via email to