evansiroky / timezone-boundary-builder

A tool to extract data from Open Street Map (OSM) to build the boundaries of the world's timezones.
MIT License
859 stars 82 forks source link

2021b and further moving away from expectations of timezone boundaries being associated to a single country #115

Closed evansiroky closed 2 years ago

evansiroky commented 3 years ago

2021b has been released and it has a number of changes that will result in significant deviations from a previous expectation (see note in https://github.com/evansiroky/timezone-boundary-builder/issues/16#issuecomment-304739065) that there would never be a timezone that would have an area in more than 1 country. Prior to the 2021b release, this project has had one exception to this rule with Asia/Bangkok which includes areas in Thailand and Vietnam. However, as of 2021b the following zone changes (as a result of some zones being moved to the backward file) would result in some zones covering multiple countries:

Zone(s) moved to backward file Zone that the area the previous zones would be merged into Countries in new merged zone
Africa/Accra Africa/Abidjan Côte d'Ivoire, Ghana
America/Atikokan America/Panama Canada, Panama
America/Blanc-Sablon, America/Curacao, America/Port_of_Spain America/Puerto_Rico Canada, Curaçao, Trinidad and Tobago, USA
America/Creston America/Phoenix Canada, USA
America/Nassau America/Toronto Canada, The Bahamas
Antarctica/DumontDUrville Pacific/Port_Moresby Antarctica, Papua New Guinea
Antarctica/Syowa Asia/Riyadh Antarctica, Saudia Arabia

So far, this project has tried to merge zones that were moved to the backward file into those that they are linked to. However, the full extent of all these merges will be large and perhaps eliciting various emotions for some people due to a timezone identifier with a city in a different country showing up in unexpected places.

As for the maintenance and boundary building process, I believe that it would still make sense to use the same OSM boundary sources from each country and combine them as needed to make these zones that span multiple countries. This process would also make it easy to "add back" a timezone from a country within a merged zone if they decide to change their timekeeping method in a way that would bring their zone back out of the backzone file.

The tzdb maintainer has been advocating for even more merges beyond these items listed above. I am opening this issue to see if there is any community input on this topic. If there isn't any major community dissent, then it seems like the logical thing to do is to keep performing the merging of timezones moved to the backward file even if the resulting merged zones now include areas from multiple countries.

evansiroky commented 2 years ago

It appears that as of 2021c, all of the zone IDs mentioned that were moved to the backward file now also have a Link entry within non-backward files. Therefore, maybe merging these zones as noted above will not be necessary.