So many specs, which one to choose?

I choose you, draft-ietf-behave-rfc3489bis-05!

The updated draft removes the attempt to discover the type and binding lifetime of the NAT device. This is a good thing as I noticed problems when testing other implementations of STUN clients where the results would be inconsistent or just incorrect, sometimes changing between STUN requests. I originally chalked the variations up to a flakey implementation but it seems the VSP (very-smart-people) working on the RFC noticed the same problem.