I like resistive over capacitive, and it doesn't matter method so much as implementation. All you have to do is try to do something precise on the iP(h)od(ne) / Storm to see the limits of capacitive screens.
I love the screen on my Kyocera Palm, but don't like the one on the Treo Pro and a few others (don't hate it, just don't like it).
Main thing will be the final device, many look good on paper and with static screenshots, but few live up to expectations.