Dangoe said:
I mean it wasn't guessed in the first place, so measured somehow - we had the uncertainties lately in the historical context. Things are still hidden from me behind a paper wall, I can ask my local uni or just pay access or try zlib, anyway at one point they were measuring, getting better, and only then we have got the unit definitions. This didn't fall from the sky.
I think about it a bit differently.
We try to measure things in the most accurate way possible. If we measure distance, we use the most accurate distance measuring tools we can make. If we want to measure time, we use the most accurate clocks we can build.
But measurement by itself is not enough. We want to be able to compare my measurement over here against your measurement over there. We need standards. Standard units. Comparisons. Reliable, repeatable, traceable comparisons. [There is a term for this: "metrology"].
We want a consistent repeatable standard against which we can all compare our meter sticks.
We want a consistent repeatable standard against which we can all compare our wrist watches.
We want a consistent repeatable standard against which we can compare our standard lab weights.
If the most consistently repeatable (not necessarily the most accurate) way we have of establishing a standard distance is to examine two scratches on a particular metal bar supported in a particular way then that is what we will use for a standard of distance. And we did for a while.
If the most consistently repeatable way we have of establishing a standard time is as a fraction of the duration of a tropical year then that is what we will use for a standard of time. And we did for a short while.
We can judge consistency and repeatability by doing repeated comparisons and seeing that they come out the same way every time within some tolerances. We can get statistics on how rough the tolerances are.
But technology moves on. If a distance measuring apparatus based on interferometry turns out to deliver more accurate comparisons than a bar with scratches while still maintaining consistency and reliability then we will change our standard definition for distance units accordingly.
But we also do not want to tie our definition to one exactingly precise description of a specific measuring device. So we do not define our standard in terms of a "
Hewlett-Packard 5710-A dual column gas chromatograph with flame analyzation detectors". Instead, we word the standard in vendor-neutral terms involving physical laws according to our well-verified understanding of how the universe works. [There is a certain amount of distaste for anchoring definitions to physical artifacts when well verified physical laws are available instead].
This, even though the whole point of the definition is to be able to realize the definition by building a device to perform reliable, repeatable, consistent measurements that match, within reason, what a hypothetical ideal device would measure.