The idea is to have a four tuple calendar:
(axial precession percentage, solation percentage, lunation percentage, day percentage)
Technically if you had an accurate enough value for the first, that's all you'd need for a very long period calendar... but I figure that having all the data there should make it easier to pin down.
The values could be expressed as floats to some standard precision.
The only problem with this is the axial precession. The others have obvious starting points: one of the solstices, one of the phases, and either midnight or noon. But I don't think axial precession has an obvious starting point?
The idea of using axial precession is that then you wouldn't have to specify which solation you're in from some epoch. Or rather, the epoch would be some part of the axial precession.
Interesting idea: perhaps it's possible to automatically determine which of the pairs (solstices, phases, midnight/noon) were used even if you pick at random, through the correlations of the whole tuple. So you wouldn't even have to standardise that!
It would probably also be possible to automatically detect when the tuple is being used. In other words, it should be quite autognomic.