> Can you refer me to any drafts/papers/emails/examples that illustrate the cons > of address selection? I have been through the archives, but apart from a lot > of arguments, I failed to find any examples that show the harm address > selection can cause vis-�-vis other methods.
I'm writing one. > > It is not acceptable to expect distributed apps to use DNS for lookups. > > DNS is too slow and too unreliable and too out of sync with the real world > > for some applications. > > Ok. If your apps are using IP addresses, then they can use the same (dest) > address (and bypass address selection) from all locations and all will work as > usual. IMHO, modifying the precedence of the prefix in the policy tables > will/should not result in parts of the network becoming "not reachable". there are two reasons this will not work: 1. if any of the addresses used are link local addresses, those will not work from a different link. (the same was true of site local addresses, but there was consensus to deprecate them) 2. with v4 networks, people are accustomed to having "private addresses" which, because they are ambiguous, cannot be routed outside of the local network. people believe they need those for security reasons. so now some of the IPv6 people are saying that since v6 does not have private addresses (nor site-local addresses, which were close to the same thing), sites will acquire unique address blocks and filter them at their borders, and set up services so that some hosts would only accept traffic on those "local" addresses. so the host or application would have to choose the "right" address in order to access those services or hosts. -------------------------------------------------------------------- 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] --------------------------------------------------------------------
