[QGIS-Developer] QGIS, DateTime and TimeZones

[Sebastian M. Ernst]

Hi all,

if you're going down the path of adding support for time zone
information, there is one thing that is not clear to me: If you are
planning to rely on "QDateTime" (i.e. Qt) for this purpose, you're
getting time zones based on UTC, but you're lacking support for UT1,
i.e. leap seconds and related issues. I know this is another huge can of
worms, but bear with me: For satellite imagery it actually matters. A
lot more imagery is taken from very low orbits nowadays, so a time
difference of a single second can - theoretically - account for up to
roughly 7 km. On the one hand side, most providers calibrate the imagery
and provide it with proper UTC time stamps. On the other hand side, the
way UTC is "smoothed" over UT1, i.e. how UT1 is correctly converted to
UTC and vice-versa, is not perfectly standardized. From practical
experience, too many people handle it in too many slightly different
ways. So when dealing with longer time series, going back to the
original UT1 time stamps (if available) is always the better and much
more accurate option. Having said that, if you want a future-proof
solution, you should definitely go beyond what Qt offers. I'd suggest to
look for a battle-tested library in this field with some astronomical
background. I guess you're going to try to solve this in C++, which
makes sense, but just as a reference, I'd suggest to have a look at
astropy's implementation of `Time`. It has saved me from serious trouble
a lot recently and serves as some sort of reference implementation for
many projects.

https://docs.astropy.org/en/stable/time/index.html

By the way, ArcGIS does not support this level of accuracy yet, too, if
you need another selling point. And since a lot of you are running on
PostgreSQL or supporting PostgreSQL-based setups, a little excerpt from
its manual: "Technically, PostgreSQL does not use UTC because leap
seconds are not handled." In older versions it ironically stated
"Technically, PostgreSQL uses UT1 because leap seconds are not handled."
It's a common theme across a lot of GIS software, which is more and more
becoming a thing.

Best regards,
Sebastian