On Tue, 28 Nov 2006 11:45:56 +0100 Stephane Bortzmeyer <bortzmeyer@xxxxxx> wrote: > > That's the problem with most "one namespace, several registries" > proposals. There is still a registry to coordinate the so-called > "several registries" so you're back to step 1. > "Most?" I'd have said "all". The name space is a tree, and (for almost all purposes except web browsing, and maybe even then) has to be one. (Reread 2826.) This says nothing about how one arbitrates the nodes of the tree at any level, including quite specifically the root node and the nodes directly underneath it (i.e., the TLDs) -- it can be the current ICANN, a replacement body, co-operation, bidding, or the ghost of Jon Postel -- but the nodes *must* be distinctly named if name resolution is to work. For those who like alternate roots, you can either assert that "our TLD registrars would never compete" (the co-operation model) or you force people trying to use these resolvers to know which one to use. In that case, the pointer to the proper resolver is the real second-level node, and the real root is whatever designates that name space. In other words, you haven't changed the model, you've just added another level (and created trouble besides). --Steve Bellovin, http://www.cs.columbia.edu/~smb _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf