It is well known that long DNA reptating in a classical gel cannot be separated under a DC field. Arrays of microfluidic posts represent an attractive alternative for performing such separations, since the larger spacing between the gel “fibers” leads to a breakdown in reptation. I will present an analytically solvable continuous-time random walk model for separations in post arrays, based on a repetitive three step cycle: (i) collision with a post; (ii) a rope-over-pulley disengagement; and (iii) uniform translation to the next post collision. For experimentally realistic values of the field, scaling laws indicate that the DNA will be in a “stem-flower” conformation when unhooking from the post, leading to a reduction in the friction when compared to a fully stretched chain. Although the stem-flower only introduces a small logarithmic correction into the averaged electrophoretic mobility, it can strongly increase the diffusivity relative to a fully stretched chain by increasing the frequency of the collisions with the posts. As a result, the separation resolution is predicted to decrease with decreasing electric field, in agreement with experiments. The quantitative agreement between the theory and experiments is satisfactory, especially considering that the theory contains no adjustable parameters.