Michael Mealling writes: > You're making assumptions that its one system. No other medium requires > uniqueness for the names _people_ use. Any medium that does not require it tends to be extremely inefficient and error-prone. > You and I are perfectly capable of understanding that there might > be two Steven Bellovins in the world. When there are twenty million John Smiths in the world, the problem becomes impossible to manage. > And conflating all of that into one system is the problem. Take those > things that humans use and separate them from those things that > computers and networks need to get things done. That's what the DNS does. But the greater the distance separating the two, the more complex, slow, and error-prone the system will be. You cannot allow human users to work in a disorderly way and expect to get an orderly result at the machine level. The system cannot think on behalf of the people using it. > Don't burden people with the uniqueness requirement when that's > not the way they expect the world to work ... They don't seem to have a problem with that "burden" when it comes to using telephones. > ... and don't burden the network with having to differentiate badly > between service behaviors given nothing but an IP address and a port > number. What's bad about the differentiation? > Again, you're conflating two different services that should be... Which > is my point. Look at the problem from a purely requirements point of > view and ignore what's been done to date..... Look at the problem from an implementation point of view and remain realistic as to what is possible if one wants any semblance of order and performance. > Reexamine the premises.... Don't fix what isn't broken. _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf