For internal addressing, why do we use DNS? In case of an address swap (hardware swap, load balancing or whatever reason), one could broadcast it to the network, and the relevant clients would acknowledge & update the address.
Why not? It's standard protocol for name-address mapping.
> In case of an address swap (hardware swap, load balancing or whatever reason), one could broadcast it to the network, and the relevant clients would acknowledge & update the address.
Client just performs DNS query before HTTP query or client caches DNS response for some time. It's solved problem and "pull" approach usually preferable to "push" approach. For example, what if that broadcast gets lost (UDP is not reliable protocol)? How application handles initial discovery?
Of course it is possible to replace DNS with something else. But why? If you're afraid that your DNS server will fail, the same could happen with your custom service discovery server.
How would you handle the ack of the broadcast? UDP can be unreliable. If you do send an ack from the destination, that singular endpoint that sent the message is going to get a number of responses equal to the number of devices on the network, which would nuke it at large scale.
