According to the CSS2 spec (see http://www.w3.org/TR/REC-CSS2/visudet.html#q17 ), floating children should not be taken into account when calculating height. That means that a box that has all children floating has the same height as an empty box (I'm not sure whether this is zero or whether it is the height of one line.) This is causing a problem at the above test (the boxacidtest). The problem is that the UL box is being enlarged to move down to the inner top (content top) of its last floating child (the LI ID="baz"). This causes the BLOCKQUOTE and the H1 to be displaced 2em downwards, 1em for their own margin and 1em for the margin of the UL. However, this movement downwards is in itself erroneous (although this part of the spec is arguable -- I think it should be changed because it leads to a layout algorithm that I think could be infinitely recursive).

*** This bug has been marked as a duplicate of 1102 ***

Verified dup of 1102 (which is now fixed).