Re: [threeten-develop] C#/.NET API
Status: Alpha
Brought to you by:
scolebourne
From: Michael D. <mic...@go...> - 2012-05-25 21:37:54
|
On 2012-05-25 16:22, Stephen Colebourne remarked on the tagging of datetimes with an indicator for the time scale from which they were obtained: >> I saw the "kind" things and thought they looked weird. Glad my initial >> judgement was right. Lets avoid a "kind" concept here ;-) >> I agree, and I'm glad to hear that. The underlying principle, I think, is that the expression for the result of a measurement (such as a length value, or a datetime) should never contain a reference to a specific quantity (such as orthometric or normal height, local time or unadjusted radiocarbon date), because such references lose their meaning at the latest when the measurement results are further evaluated (eg, by computing a variance, or by computing a rate through dividing measurements of two quantities). This principle is also propagated by the BIPM, the international gremium for the standardization of measurement units. Nevertheless, it is often violated: the IPCC use kg[C] for kg of carbon as if it were a unit, electircal engineers talk about V[pp] when they mean twice the amplitude of a signal measured in volts, and computer scientists use 2012-05-25T20:45:15Z as a datetime value, implying that it is a reading of UTC. While all these notations may have their use as shorthands (eg, for "UTC at that moment was 2012-05-25T21:37:15"), one has to realize that the underlying measurement values are just masses, voltages, datetimes, without any adornments. All the operations on the values are defined independently from where the measurement came -- the result of the operation 2012-05-25T21:37:15Z - 05 h may signify UTC five hours earlier, or it may signify a local time, or it could even be a reading of Martian time scale that is currently 5 h behind UTC (such time scales for Mars, to be used with the Gregorian calendar, have been proposed). Any assumption by software about the meaning of the result is at best useless (if the software ignores them), and at worst annoying (if the result of an operation of the software may depend on such assumptions). Michael Deckers. |