[Haskell-cafe] Bug-free, leak-free, battle-tested library for broadcast channels?

[Joachim Durchholz]

Am 25.01.2017 um 09:47 schrieb David Turner:
> The trouble with the real-time clock is that it can change
> discontinuously, even backwards, and it is hard to account for things
> like leap seconds with it, so durations calculated by subtracting two
> UTCTimes are a little unreliable.
>
> OTOH you might well want to know the real start and end times of your
> process instead/as well - it depends on your application.

You use the proper tool for the job.
Monotonic clock for measuring durations, UTC for reporting approximate 
points in time. I.e. something like
   2017-01-12 10:43: benchmark #3 started
   2017-01-12 10:43: benchmark #3 took 3.472 ms to complete

Mixing the monotonic durations and UTC is a path to madness, and nothing 
good can come from it. (SCNR the hyperbole.)