3 ms·
RFC 3339 should be preferred to iso, but it does not solve the wall clock issue. In a calendar application I worked on, we use city name and lat/lng + "naive" t
by kuon 2y ago
RFC 3339 should be preferred to iso, but it does not solve the wall clock issue. In a calendar application I worked on, we use city name and lat/lng + "naive" timestamp (yyyy-mm-dd hh:mm or moon based date or other specialties like Japanese year depending on the locale). And we "resolve" to UTC + offset when the date is close.
https://news.ycombinator.com/item?id=37339535 https://news.ycombinator.com/item?id=37339535
https://www.mooncratertycho.com/the-12-lunar-calendars-still-in-use-around-the-world/ https://www.mooncratertycho.com/the-12-lunar-calendars-still...
- 8n4vidtmkvmk 2y agoNice! https://i.imgur.com/vsC2Wyd.png https://i.imgur.com/vsC2Wyd.png See even RFC 3339 seems unnecessarily lax. You can use _ or lowercase t or space as the date and time separator? Why? Just pick one instead of complicating all the parsers. The intersection of the two specs seems pretty good though.