@jackie_w: sure. I imagine that there are cases where millenium makes the most sense, but I suspect they are few and far between.
The question is whether or not to really do something like the numbers, that is use different partitioning depending on the value. Dates can run from 01/01/0101 to far in the future. My suspicion, dealing with books, is that the dates around "today" are of more interest than, for example, 1729 or 2350. Perhaps something like:
- Group by decade for 50 years in the past (future), rounded down (up) to the century.
- Group my century to the nearest larger/smaller millenium.
- Group what is left by millennium
Assuming today as the reference date, we would see:
- 1/1/0101 to 31/12/999 ==> 0000-0999
- 1/1/1000 to 31/12/1889 ==> group by century (1000-1099, 1100-1199, ..., 1800-1899)
- 1/1/1900 to 31/12.2099 ==> group by decade (1900-1909, 1910-1919, ..., 2090-1099)
- 1/1/2100 to 31/12/2199 ==> group by century (2100-2199, 2200-2299, ..., 2900-2999)
- 1/1/3000 to whatever ==> group my millennium (3000-3999, 4000-4999, ...)
The leading zeros won't be displayed on the first range.
The "50" is arbitrary, defining the block of years centered on "today" where "first letters" are decades. Any multiple of 10 will work. I think it should be smaller than 100, but am willing to argue.
And yes, I know that a millennium runs from 1001 to 2000 not 1000 to 1999, but the calculations are easier in the second case. It might also be visually easier to use.
Thoughts?