I sort of agree. The addressing, a naming function and routing are the three legs. If you do naming right, you can drop addressing and use ephemeral addresses, and if you do routing right you can drop addresses and do ad-hoc. But you need addresses and routing if you want to do without names, so I kind of see this as a triumverate we're grown into now (obviously there are other legs, like security, which we always cut away in launch and then miss. three legged stools are not stable...)
But that said, I see this other dimension. we don't run r* commands any more. We don't run UUCP any more. Things which feel baked in turn out to be ephemeral to the core function. So name functions? Up where Brian Trammell is writing drafts? Sure. we need that. What underlying protocol it maps to, Thats a big statement. I don't buy that forever more amen it maps to UDP. If somebody makes DOI work over ICMPv6, I could believe in 25 years we'd migrate to bootstrap of DOI via ICMPv6 and be out of the DNS moment entirely. As it stands, almost all bootstrapp-y application phases accept address literals sorry [address:literals] somehow. Names are only a convenience function, set against routing. _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
