Also, NTP runs on UDP rather than TCP, and it only requires one packet from each party to exchange time, so it's harder to rate-limit it without making your server unreliable.
Because that's the point of NTP; internally both operating systems use UTC, and in practice most clients can be in sync to within 1 ms or so of the true time, given reasonable connection speed and quality.
1. https://d38if4m2in2lkc.cloudfront.net/2016/09/the-school-for...
2. https://d38if4m2in2lkc.cloudfront.net/2016/10/the-school-for...
3. https://d38if4m2in2lkc.cloudfront.net/2016/10/the-school-for...
4. https://d38if4m2in2lkc.cloudfront.net/2016/10/the-school-for...
5. https://d38if4m2in2lkc.cloudfront.net/2016/12/the-school-for...
I probably won't read (m)any of the comments below, but if I had to pick the "do not miss" parts, they would be https://d38if4m2in2lkc.cloudfront.net/2016/10/the-school-for... (It really isn't a typical consensus algorithm.) and https://d38if4m2in2lkc.cloudfront.net/2016/12/the-school-for... (Them: "1 NTP peer is better than 2"; me: "Don't make me come down there"), but really, you should just go read https://tools.ietf.org/html/rfc8633