Time math is ambiguously-specified

[davidben]

Leap seconds, UTC, TAI, etc., are all the worst. While we never actually write the timestamps on the wire, it is important that everyone agree, for a given batch number, what times they refer to.

The intent was that you just do math in POSIX time, since that's what every programming language gives you. But if there's ever another leap second, it gets a little tricky. It might be that we have to literally cite POSIX here? Something like: all of these are just measured in seconds, do the math in seconds to get a POSIX timestamp, convert the current time to POSIX time, and then compare in that space? https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap04.html#tag_04_16

Otherwise you have to answer ridiculous questions like what happens at one of the POSIX timestamps that correspond to multiple times of day.

Not that anything at this granularity matters at all for a PKI.

[davidben]

Notes to self for specifying this:

• C++ says that std::chrono::system_clock says "time since 1970-01-01 00:00:00 UTC excluding leap seconds"
• POSIX has a DOI, so generating the citation is pretty easy.