On 2021-05-23 02:33:22 (+0800), Paul Eggert wrote:
On 5/22/21 11:17 AM, Tim Parenti wrote:
In particular, the latter group (things like America/La_Paz -> Etc/GMT+4) seems to encourage things or behaviors which historically cause confusion, especially for novices:
We could avoid these problems by merging Etc/GMT+4 into America/La_Paz rather than vice versa. That would be easy to arrange, and would result in the same number of timezones. A downside would be that TZ='Etc/GMT+4' would no longer be equivalent to TZ='<-04>4' for pre-1970 timestamps, but those timestamps are out of scope anyway.
While those timestamps are out of scope, software exists that is aware of tzdb names but not the intricate details of POSIX TZ strings. There are use cases where being able to set a fixed-offset timezone is desired. Breaking the current behaviour for pre-1970 timestamps would not be cool. I'd even suggest documenting explicitly that Etc/* are aliases for fixed offsets.
if any of the resulting pairings are particularly far-flung (I haven't checked thoroughly yet)
Some are far-flung but that should be OK. It shouldn't matter, for example, that Asia/Dubai and Asia/Reunion are aliases even though Dubai and Réunion are separated by 5000 km of ocean and disparate political systems. Users of those aliases shouldn't care, any more than users of Europe/Copenhagen and Europe/Berlin should care.
(I assume Indian/Reunion. :)) I agree that this should not be a concern. Philip -- Philip Paeps Senior Reality Engineer Alternative Enterprises