[tahoe-dev] timestamps (was: collective effort to improve logging)

Michael Walsh michael at michael.ie
Tue Sep 18 18:17:43 UTC 2007

On 9/17/07, Brian Warner <warner at allmydata.com> wrote:
> Also, it turns out that TAI64N only has nanosecond
> resolution: the format is hex(seconds)+hex(nanoseconds), and it omits the
> decimal point. You have to use TAI64NA to get attosecond resolution.

Do you think it's actually useful to get that sort of resolution?

I'm a big fan of UTC time - I standardised my logs (and how times are
displayed to the end user) of a project on UTC a long time ago. It
worked out quite well.

So far, have you ever needed more than to-the-second accuracy?
Milisecond accuracy would be interesting to have if benchmarking
function calls.

Then there is the issue with actually loosing precision. I've got no
idea how accurate python timings are, but I do know that with Java it
can be very very difficult to get an accurate high resolution timer.
There is no point in storing values that aren't accurate.

If to-the-second is all you want, we might aswell throw EPOCH time
into the list (doesn't solve timezones/notation though)


More information about the tahoe-dev mailing list