Eliot Lear wrote: > Having dipped my toe in, I'm going to bow out of this debate after two > comments: > > First, by making the IP address of the root server part of the name, you > have merely moved the problem, and not created "multiple roots". The IP address of the root server is part of any name today, and it resides in your browser. To contrast, what I said does not have the IP address of the root server as part of the name, since a name can now be tied to any root server a posteriori. Which root server to choose is decidable by the sender, and this decision can be rather easily automated as I mentioned yesterday, including cache. Late binding versus early binding. > The > root is now the routing system, with all of its assorted warts. Worse... The root is part of the routing in a hierarchical tree. > >> The way things work today, there isn't one single > >>address for the IANA root. > > > > > > which supports my point. > > One IP address, one point of failure within the topology, unless you can > play games with anycast. Is that what you're proposing? I don't understand your question. I was merely referring to the fact that we prefer to use multiple IP#s for the root servers even though there is no perfect answer to synchronize them (race conditions set in, for example). Cheers, Ed Gerck