That was the original definition. Things change over the course of a couple of hundred years or so, though, and the imprecision of that definition was noticed early on.
Where does the imprecision come from exactly? Can't we derive the kilogram from the number of hydrogen and oxygen atoms in a litre of water at an uniform temperature?
But then your definition depends on the definition of temperature (energy/entropy) which in turn depends on the definition of mass. You might say "make all measurements at the triple point" but then you are still stuck with the problem of dealing with thermal fluctuations and structural variation (liquids have uniform structure only in a weak statistical sense) in addition to the practical problem of ensuring uniform composition at the triple point. This idea doesn't work for benchtop measurements (How do you determine that you have 1 mole of H2O? By measuring it's mass!) and it doesn't make sense for high-precision measurements where you need to provide a mechanism to allow the experimenter to make the error arbitrarily small.
One of the leading proposals for the 2014 redefinition is very similar in spirit, though: defining Avagadro's number to be an actual number rather than a derived quantity (effectively this uses Carbon 12 as the mass standard).
Another proposal is to define Planck's constant. I can't comment on the measurability of this one, but it would probably make people studying atomic physics happy because they like to use units that set most common constants to one, and the redefinition would make more of these implicit factors of 1 defined, rather than measured quantities.
Perhaps. But at least one complexity there is that there are a couple of isotopes (of different mass) for both oxygen and hydrogen.
There's the Vienna Standard Mean Ocean Water[1][2] which accounts for differing isotopic compositions in different parts of the world.
There's been some competing attempts to the Standard Kilogram, including a pure silicon sphere of extremely precise physical dimensions (which we're good at), and using the lattice spacing to determine the number of atoms.
