I don't want forwarding. If I move from California to New York I want mail sent to me from New York not to have to traverse the country twice to get to me.
I think those services exist in some countries, but not from postal services themself, so delivery-time is still rather slow.
That's the idea: the redirection would be handled by the carriers, not some third party. In place of the current system of (a) scan address; (b) route package; (c) deliver package the process would be (a) scan recipient ID; (b) translate ID to address; (c) route package; (d) deliver package. The address lookup should be completely automated and introduce no discernable additional latency in the mail delivery. It could even be implemented as a special case of the address normalization step carriers already perform.
In the first place there is not even enough data to know whether the huge investment would be justified.
Or do you have one address for UPS deliveries and a different address for FedEx deliveries? (Substitute local carrier names as appropriate.) While that's more work—and pointlessly complex IMHO—you'd just need to set up one ID per carrier and give the correct ID, as opposed to the carrier-specific address.
Anyway, a shortcut to implementing indirection for these less-organized postal services you're describing would be to look up the recipient's ID when the parcel first enters the system and affix a label with the full address. That way the sender doesn't need to know the address and everything else about the delivery can stay the same.
In that scale, nothing is simple, because everything can become expensive on the high run.
Only if you don't consider flexibility part of the task, which is rather short-sighted in today's world. Maybe postal services could afford to be inflexible once, but those days are long past.