We had a job like this come through as a proposal a couple months ago. Guy had a salesman in zip code A, B, and C. Had a customer in zip code D,E, and F. Wanted to know which salesmen was closest.

The results would be stored on his server. He was basically expecting a hash of each Zip Code centroid in the US measured to every other zip code centroid. Hate to bust the guy's bubble but it would've taked years to do all the calculations, assuming really fast calculations.

So.... Maybe you could eliminate all the zips that aren't + or - 1 utm zone from the one you're measuring from. That might eliminate quite a few. Maybe speed it up a bit. Neil