> >=> This makes a lot more sense IMHO. How can we write > >"MUST NOT use site-local when global addresses are available" > >in a spec? I mean how can we enforce that? If we can't enforce > >a "MUST" or "MUST NOT" then it shouldn't be in the spec > >in the first place. > > I disagree. The scoped addressing architecture will define how > scoped addresses are used on the network. It already has lots > of things in it that can't be "enforced" programmatically on a > single node, such as: > > - scope boundaries are in nodes, not on links > - scopes are contiguous and convex > - smaller scopes are completely contained within > larger scopes > > All of these things require that an administrator configure a > network in a way that meets these guidelines.
=> Because if he/she doesn't do that, communication will fail in many cases. But how can one force administrators who (for some reason we don't know now) decide to use both global and site-local prefixes in a site and defined SBRs to support this? That would break a MUST NOT in the spec, but things will still work. > > Similarly, we could say that site-local addresses should not be > assigned on globally-connected networks. > > Besides, it would be possible (although perhaps not adviseable) > to enforce this restriction. Nodes could immediately deprecate > site-local addresses whenever a global address is configured. => Just so they feel that they followed our standards? I don't think so. A much better approach IMHO would be to highlight the problems and discourage people from using site-local addresses for globally connected sites _because_ of the highlighted problems. After that you can hope that people will follow. That's all we can do. Hesham -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
