Re: Exposing subsecond precision in current-seconds

[Lassi Kortela]

> Sorry, I meant to write microseconds.

No problem. In that case, please take a look at Gambit's time API: 
<https://www.iro.umontreal.ca/~gambit/doc/gambit.html#Measuring-time>. 
It uses a float for the internal representation of time objects. 
`time->seconds` extracts it an ordinary float out of it. Perhaps the 
same interface could be used in Chicken.

> For nanoseconds precision you'd need to
> use `clock_gettime` which is harder to emulate on Windows than `gettimeofday`.

The main point of nanosecond precision in Scheme is to preserve 
bit-for-bit precise timestamps from Unix file system APIs (in case you 
copy the timestamp elsewhere and somebody later compares it for exact 
equality with the original). NTFS timestamps have 100-nsec precision, so 
a (seconds nanoseconds) pair is good for them as well.

It's unlikely that useful ns-level accuracy for the current time is 
obtainable in any Scheme program (or any program generally). System call 
overhead is measured in microseconds. I guess CPU cycle counters in 
theory have nanosecond accuracy, but those are useful mainly in 
computing the time elapsed between two counters, not in reference to 
exact moments on a standard timescale like UTC.