This thread nerd-sniped me into looking up the exact definition of the dimensions. Neither standard gives the formula used to obtain the specified sizes (for A4, precisely as above, ±2 mm). But what's interesting is that we can tell from the rounded values that it is _not_ the geometric sequence that the stated principles imply,

  x(n) = 1000 * 2^(-1/4 - n/2)
  y(n) = 1000 * 2^( 1/4 - n/2)

y(0) = x(-1) = 1189,2 would map to the 1189 mm long side of A0 and short side of 2A0 in the standard; y(1) = x(0) = 840,9 to 841 mm; and y(2) = x(1) = 594,6 to 594 mm -- inconsistent with any common rounding scheme. This also rules out evaluation of the recursive form with intermediate rounding, and/or starting with the approximations 1,414; 0,841; and 1,189 as written in the standard.

Instead, a form that _is_ consistent is:

  x(0) =  841 = round(1000 / sqrt(sqrt(2)))
  y(0) = 1189 = round(1000 * sqrt(sqrt(2)))
  x(n) = floor(y(n-1) / 2)
  y(n) = x(n-1)

Which makes sense - once A0 has been calculated, only halving and doubling are needed to determine the other sizes.

B0 is even easier to memorize: 1000 mm x 1414 mm