James disagrees with my feelings on using UDDI for transport-agnostic rollover: "Sure, if it wasn't for the total lack of reliability, security, and less than stellar performance. IMHO, UDDI is too fat for serious runtime fail over. I mean think about it, if you're a manufacturer and you need to order some parts for immediate shipment, but you can't get a hold of your favorite supplier, are you going to go trudging around through the yellow pages looking for a replacement or are you going to already have a replacement lined up and programmed into the speed dialer on your phone? It just doesn't make sense to wait until something fails to go out discover a replacement service."

I agree. Using the public UDDI for roll-over doesn't make sense. I mean using UDDI within the enterprise and behind the firewall. Using the public UDDI servers doesn't make sense for performance specific operations.

In general, I think the scenario of using UDDI to find partners to interact with is overblown, or at least years away. I don't think those scenarios warrant UDDI, or any programmatic approach. (Maybe someday in the far, far future.) For my purposes UDDI is for programmatic discovery of specific endpoints that implement specific tModels where you already have an existing relationship - usually behind the firewall, or (in some cases and much more rarely) in established B2B relationships.