I like the dead parrot reference. So we have already discussed that no one is doing dynamic runtime binding to services or discovery. Most organizations are not going to have multiple services doing the same thing. A clarification here, this is not to say organizations do not duplicate work unintentionally but rather I am talking about writing multiple services doing the same thing on purpose. So for organizations with a small number of services, UDDI probably isn't all that necessary. I'm not saying it would not be useful just not necessary.
Of course as services grow, the ability to manage them diminishes. Something like a UDDI based repository would come in very handy at that point. I'm still not convinced about the other stuff such as SLA management, failover, dynamic discovery etc. It sounds pretty cool but I'm not sure how many companies will really need that. It probably has a bigger role with external service providers.